5% THD to .000 THD SOUND

This is a complex topic. But if you understand how the ear works it gets easier. So let's start there.

Here are the human hearing/perceptual rules that are most important regarding this topic. There are three. They are:

1) the human ear/brain system used higher ordered harmonics to determine sound pressure
2) Nearly all forms of distortion (harmonic, IM and inharmonic) are translated by the ear/brain system into tonality
3) The ear/brain system has a complex series of tipping points; one of them being that it can easily favor tonality caused by distortion over actual frequency response variation.

The implications are immense. Now the other important thing to understand is that most of what we know about human hearing physiology has occurred in the last 40-50 years.  

Our test and measurement regime OTOH is based on hearing knowledge dating from about 1960 and ignores most of the research of the last 60 years.

So finally we can discuss distortion in amps. It should be obvious now that distortion is far more audible than the older textbooks might have you understand. The lower ordered harmonics contribute to 'richness', 'warmth', 'bloom' and similar audiophile comments. The higher ordered harmonics (in trace amounts due to the 1st rule above) contribute to 'brightness', 'hardness', 'brittleness' and similar audiophile comments. 

Usually loop negative feedback is used to reduce distortion in amplifiers and also has the secondary aspect of controlling output impedance. The primary reason for using it is to guarantee flat frequency response on a variable speaker load (the effects of rule two are ignored). This concept was developed by MacIntosh and EV in the last 1950s and early 1960s. By 1975 this idea had dominated the industry. I call it the Voltage Paradigm as the amplifier has to behave as a voltage source in order for this to work.  The prior art is known as the Power Paradigm and is still in limited use today (SETs, horns, planar speakers and generally tube amps with little or no feedback are examples).

The problem is feedback is known to not only suppress lower ordered harmonics but the price paid is higher orders are added, in many cases clear out to the 81st (ref.: Norman Crowhurst). In addition, intermodulations can occur at the feedback node, generating IM and inharmonic distortions (IM and inharmonic distortion are sometimes the same thing, sometimes not- both contribute to brightness).

Amplifiers that are linear enough to operate without loop feedback often do not contain much in the way of higher ordered harmonics. SETs (up to about 20% of full power) are a good example, although they do seem to represent a special circumstance worthy of their own thread). If feedback is applied, the higher orders will be present even at lower power levels (and in an SET the distortion is often unmeasurable at lower power levels).

The second harmonic often attributed to tubes is not something inherent in tubes (transistors actually make more). It is really more a function of the topology of the circuit. You can avoid the second harmonic by building an entirely differential/push/pull circuit and this is how most transistor amps avoid it. This can be done with tubes as well.

The ear does not care too much about the lower orders (2nd 3rd and 4th) but it seems to care a lot about the higher orders (5th and above). Its been known since the 1930s that the 7th contributes to a metallic quality in the sound, even in small amounts. So you can have a fair amount of distortion (5% at full power) and if its all lower ordered harmonics the amp might sound just fine. OTOH 0.01% of only higher orders is going to come off more likely as a bright amplifier. This is why two amps on the bench can have the same bandwidth but one might sound bright and other not.

IM distortion, unlike THD can be objectionable even in small amounts. This is also true of inharmonic distortion (an example of that might be intermodulations associated with a scan frequency, which might occur in a digital system, a SMPS or a class D application, although the latter two seem to be fairly sorted out now as opposed to 20 years ago!).

Now if its not obvious by now, since the understanding of how the ear/brain system works is pretty well ignored by the audio industry, the result is that the spec sheets often wind up being a good example of the Emperor's New Clothes. IOW they make the amp look good on paper but lacking a correlation in sonic performance.

It is the contention of the Power Paradigm proponents that the tonalities generated by distortion often play a bigger and more objectionable role than do actual frequency response errors, which can be managed by careful amplifier/loudspeaker matching.

The tube/transistor debate seems to revolve around the aspects of this conversation. The bottom line is that there really isn't a simple answer and for that reason audition is often the best solution.  If you find that you have a preference (for example, many people like tubes, which is why this so-called 'obsolete' technology is still around half a century after being declared so; obviously the market wants them for some reason and so is keeping the tube industry in the black) then you will also find that certain speakers favor certain amplifiers and thus equipment matching is simply an issue that can't be ignored. There is no one amplifier, tube, solid state or Class D, that can play well with all speakers. For more information see:

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

In the end, you alone will have to sort out how to approach the issue of creating a system that is at once musical and neutral.

Atmasphere, I'd like your opinion on Bruno Putzey's latest class D amp design philosophy.  I've read that he once held the same view as you do regarding negative feedback producing random upper order harmonics, but he's now broke with that orthodoxy.  First, he claims that negative feedback needs to be evenly applied across the frequency spectrum and not disproportionately at lower frequencies.  Second, Putzey claims that as you apply more and more negative feedback the odd harmonics disappear and distortion levels are low and smooth.  I believe he is talking about 40 to 60 dB loop gain.  Here's a link to Putzey's Kaluga amp's measurements.

What you've written makes a lot of sense to me about why some amplifiers sound the way they do, but Putzey seems to be going in an opposite direction and his designs are said to be aurally excellent.

Ralph,
I’d say that your description of the sound of amplifiers correlates very well to my own listening experiences over the years. I’d also be interested in your reply to Onhwy61's inquiry. I’ve heard the Putney amplifiers twice (under show conditions) and was underwhelmed. It could be due to the show environment but there were other rooms at the show that had wonderful sound so who knows,
Charles,

Atmasphere, I’d like your opinion on Bruno Putzey’s latest class D amp design philosophy. I’ve read that he once held the same view as you do regarding negative feedback producing random upper order harmonics, but he’s now broke with that orthodoxy. First, he claims that negative feedback needs to be evenly applied across the frequency spectrum and not disproportionately at lower frequencies. Second, Putzey claims that as you apply more and more negative feedback the odd harmonics disappear and distortion levels are low and smooth. I believe he is talking about 40 to 60 dB loop gain.

There is not a simple answer to this. First, to be clear, I don’t agree about the bit of ’random upper order harmonics’; they are a bit more predictable as you will see below. Also, I think you run into some classic issues with gain and phase margins if you do as Bruno proposes (see Norman Crowhurst; this book is an excellent primer, look on page 3-15 and read from there: http://www.tubebooks.org/Books/crowhurst_basic_3.pdf)

I’ll go with Norman Crowhurst any day- math has a way of not going bad as it gets older.

There is more to loop feedback though. One of the bigger problems that is rarely dealt with is the issue of Radio Frequency Interference (RFI). The speaker cable tends to act like an antenna, and since the feedback loop takes information from the output of the amp and feeds it back into the input for correction, RFI is also thus injected into the amplifier.

This can cause havoc- RFI is not helpful in audio circuits! It might be why an amplifier might sound really bright in one situation and in another with the same speakers, might seem pretty good. RFI can be an unknown variable and susceptibility.

Yet you rarely see feedback loops that deal with this problem.

Bruno is correct that as you add more feedback you can eventually decrease the higher ordered harmonics. However, as Nelson Pass has pointed out in this excellent article on feedback:

https://passlabs.com/articles/audio-distortion-and-feedback

Quoting from that article, Nelson puts the problem in a nutshell:

Negative feedback can reduce the total quantity of distortion, but it adds new components on its own, and tempts the designer to use more cascaded gain stages in search of better numbers, accompanied by greater feedback frequency stability issues.

The resulting complexity creates distortion which is unlike the simple harmonics associated with musical instruments, and we see that these complex waves can gather to create the occasional tsunami of distortion, peaking at values far above those imagined by the distortion specifications.

If you want the peak distortion of the circuit of figure 13 to remain below .1% with a complex signal, then you need to reduce it by a factor of about 3000. 70 dB of feedback would do it, but that does seems like a lot.

By contrast, it appears that if you can make a single stage operate at .01% 2nd harmonic with a single tone without feedback, you could also achieve the .1% peak in the complex IM test.

I like to think the latter would sound better.

As we know class D reacts the same way to feedback that any other amplifier does, being an analog process. Generally its been my conclusion that while Bruno is right in theory (except where he isn’t, see Norman Crowhurst), he’s wrong in practice.

Atmashpere, thanks for taking the time to reply.

I misrepresented Putzkey when I said the distortion spectrum resulting from negative feedback was random.  I should have said higher order harmonics instead of random.

Whether is right or wrong in theory or in practice is being tested in the marketplace.  Theta, Bel Canto, MBL, Jeff Rowland and others are using Putzkey's amplifiers in their products and they are getting positive reviews and user feedback.  I guess only time can really sort these things out.

^^Well if he has a lot of gain (and a lot of feedback), has kept RFI out of the feedback loop and is paying attention to phase margins, it should work.

At the onset of clipping, the clipping behavior will not be graceful- that can't be avoided. That is where zero feedback really shines. But if clipping is avoided and the feedback variables are paid proper attention, the result should be excellent.