• Recently Written

  • How to Bias Your Amplifier
  • Output Tube Biasing
  • DIY 6T9 Push-Pull Tube Amplifier
  • FOSTEX FX120 ML-TQWT SPEAKERS
  • DIY Speaker Cables
  • Building an Output Transformer for a Tube Amp
  • Build a DIY vacuum tube amp (EL34 / KT88)
  • Amp Dates, Tube Charts, Transformers and Information
  • Vacuum Tube Amplifiers
  • Buy A Guitar Amplifier For Pure Volume!

• Categories

  • Amplifier
  • Articles
  • Speakers
  • Transformer

• Archives

  • December 2008
  • November 2008
  • October 2008
  • September 2008
  • August 2008
  • July 2008
  • March 2008
  • December 2007
  • November 2007

• Admin

  • Log in
  • Wordpress

  Search

  
  
  

Since 1999 when the original article “Why SE Tube Amplifiers?” was placed in the Internet I have received some feedback from readers. Part of this feedback has been in the form of direct questions and some through discussions on Internet groups that I have either followed or that I have been called attention by someone. Several comments either pro or against the ideas have been raised. Except for some of the questions I have personally received I have never had time to directly answer anything about it, therefore I feel that about four years later a short list of the more relevant and/or recurrent comments could be prepared. I also felt that some of the criticisms to points totally unrelated to the main issue of the article that have appeared at Internet groups also deserved an explanation.

Why was the article written?

During the investigation about the output impedance of single ended tube amplifiers that had its results published as SE Amplifier Output Impedance (Glass Audio 3/97 and 6/97) it was very clear that, as expected, the effects it produced could have some major influence on the audible results. Those articles discussed these effects and have shown that under certain circumstances it could result in a performance of the system composed of single ended amplifier and loudspeaker that could “objectively” (this meaning when evaluated by the standard measurements as they are made at this point in time) be considered better. The fact that these circumstances were not very common does not rule out the possibility that they could happen either by chance or by deliberately designing the system to take advantages of them. These two articles dealed with why and under what circumstances a 3 ohm output impedance could not only still sound good but it could actually have the potential for a better audible performance than a perfect amplifier with 0 ohm output impedance.

The Why SE Tube Amplifier article was written with the same idea. To describe what was found during the exploration of how a 1% THD rating at 1W, for example, could result in the potential for better audible results. And also describe under what circumstances it could also have better “objective” measurements.

The original article is quite long and I believe some of the criticism it received is from people who have not read it entirely and have concluded that it was just a piece to defend that single ended tube amplifiers are “better”, when its real initial goal was to discuss how they could sound very good to some people to the point that these amplifiers were their preferred alternative and under what circumstances those “awful” THD ratings could either not be a problem or even be beneficial.

How do you account for the psichoacoustic factor?

When I have called attention to the fact that there may happen a reduction of measured THD when connecting single ended amplifiers to loudspeakers and that this reduction has a much greater probability of happening than is normally acknowledged I have left an open door to take in account the psychoacoustic factors involved in not hearing some types of distortion in amplifiers. In fact I have explicitly said: “…I agree that psychoacoustic tests may find that some kind of distortion is good or may improve a particular aspect of the reproduction, but I believe that in the long run we all look for less distortion. But, in any case, even if we are looking for some distortion effect, everything that I will say will still hold true and may even help us to achieve this target…” . I have not read anything that shows to me with any reasonably founded evidence to what extent these psychoacoustics ideas are true (and this can be because I just could not find the right papers) neither I have experimented enough with it to be able to say anything relevant about it. The only thing I know is that whenever those very pretty distortion spectra of SE amplifiers with decreasing levels of harmonics interact with the loudspeakers own distortion the resulting final distortion will only look similar if the amplifier distortion is much greater than the loudspeaker distortion …and this is not the case most of the time. Therefore I believe that to reach psychoacoustic conclusions based only on those type of amplifier distortion spectra is certainly risky.

The distortion of loudspeakers

Loudspeaker distortion is referred to in my article and some just complains were heard about the following quote: “…if we examine the distortion spectrum of any reasonable speaker we will see that all the distortion is low order. In most cases it is mainly second harmonic with some third harmonic also appearing with lower levels. The higher order terms will only become noticeable near the limit of power.” We must understand the context of this quote. Loudspeakers do produce higher order distortion products but compared to the amount of second and third order distortion it may be considered much less significant. Still the main point is that the levels and distortion spectrum resembles the distortion levels and spectrum of single ended amplifiers that also do produce higher order harmonics as the level is raised. Another interesting issue is that simpler loudspeaker constructing practices will usually produce a distortion spectrum that is more closely related to typical SE tube amplifiers. Whenever more sophisticated magnetic structures or compensating suspensions are employed the result is a reduction in the THD readings of the speaker but this is achieved usually reducing mostly 2^nd order distortion, sometimes also the 3^rd so the spectrum may be changed.

Measuring loudspeaker distortion

One of the most recurring questions has been about how the measurements shown have been taken. Measuring loudspeaker distortion and arriving at a simple number can be a very difficult task. Sometimes slight changes in the relative position of speaker and microphone can produce significant changes in the result. The ambient where you measure it can also have a very definite impact specially if you are using a sine sweep. A sure way to make repeatable measurements would need an anechoic chamber and very precisely defined relative positioning. The purpose of the measurements shown was to show the reduction of distortion that may happen with typical single ended amplifier connected to a loudspeaker. Therefore this is a relative measurement. I have taken all the three measurements without touching the positioning of the speaker or microphone. I have used a sine sweep in a relatively large (about 100m³) room that is fairly dead acoustically but it is very far from anechoic. Every care was taken so that the microphone own distortion would be small enough to not be a factor in the measurement. But still this can only be viewed as a relative measurement and the distortion graphs are only relevant to show that what was calculated may really happen.

Most of the criticisms to the article are based on the fact that only one case is shown. The purpose of the graph is to show that this phenomenon actually happens. The probability of it happening can be deducted from the math and is stated very clearly in percentile terms. Actually if the 2^nd order distortion of the amplifier is less than that of the speaker (but still significant) the probabilities of distortion reduction will even increase although the amount of reduction will decrease. Even hundreds of measurements could not provide the insight that these simple calculations give us if we can really understand them. No limiting assumption has been made related to the fact that 2^nd harmonic distortion is generated by electronic or mechanical means. There should be no need for anything else to understand that this is a very real thing that may have a huge significance. Also this article is not a graduate student thesis where there is the need to show every little detail. If someone can point to a flaw in the mathematical reasoning - that is in fact quite simple - then I will not only pay attention but I will be glad to review my position which is that this effect is a very significant one that has been grossly overlooked.

Some points about distortion reduction

A very interesting comment addresses the fact that the distortion of speakers will vary with power. For the reduction of distortion to be significant there is a need for the variation of the amplifier distortion with power and that of the speaker to happen somewhat similarly. This may indeed look difficult but it just seem also to happen more frequently than we would normally guess. There is in fact some reasons for it to happen often but I will save this discussion for a projected sequence to the original article. Of course the fact that the probability of reduction is far greater than expected also helps.

Another issue raised is that looking at the three graphs of the distortion against frequency it looks like the reduction in fig.3 is much more apparent than the increase in fig.2. This is true at first sight but we have to consider at least four reasons for this to happen. First the graphs plot the percentage of distortion in a logarithmic way. The difference between 1% and 2% looks larger than the difference between 2% and 3%. This is the standard way of plotting level increases. The second issue is that there is actually the possibility that the reduction is larger if you look very carefully. The same logarithmic way of plotting would show that the maximum increase in distortion would be seen as doubling its value while the maximum reduction would reduce distortion to zero, and this would need an infinitely extending downward expanding paper sheet! One interesting comment is that this logarithmic representation of levels is believed to express better how we hear. The third issue is the higher output impedance of the SE amplifier. Higher output impedance may alter, usually for less, the distortion levels of a loudspeaker. This is a totally different issue and a much more deep discussion is needed to clarify this point. Now I can only point that most of the experiments about output impedance and distortion levels were done simultaneously and it is fair to say that they are somewhat interrelated in several ways. Dividing the results in two parts contribution of the high output impedance and of the high THD readings is important to make it possible to understand the different aspects but it is also important to keep in mind that these two aspects influence each other. Finally the fourth reason
that can explain certain unexpected differences - although I am sure this is not a relevant issue in this case - is that the distortion produced by the amplifier is dependent on the load. Reversing a load like a loudspeaker may result in slightly different distortion levels. This is so because there are small asymmetries in the dynamic behavior of the load depending on the connection.

Another issue is what happens with the 3^rd and higher harmonics. In theory there always would be a way to reduce or even cancel all distortion no matter what harmonic. The practical aspect of what can be done is a different matter. Also a very simple method as using a single ended tube amplifier to drive a speaker is even harder to be seen as an effective way of doing it. But this does not mean that we should automatically assume that the higher order harmonics will simply add as 2+2=4 in all circumstances.

The real definition of Class A

There were criticisms about my assertion that “…A single ended output stage has to work in pure class A always. There is no way to make a single ended output stage work in any other class (AB or B). These other classes of operation require, by definition, that we switch the conduction from one device to another and in a single ended output stage there isn’t another device…” . May be the word definition can only be used in a very strict sense in English and sometimes this is not what happens in my native language. Whatever it is, I apologize for any inconveniences it may have created but the article main point certainly has nothing to do with this. A broad definition of Class A in most textbooks usually states that a device is operated in Class A whenever there is conduction over all the cycle. But under the somewhat smaller audio context if we look at Radiotron Designer’s Handbook page 545 you will find: “Class A operation is the normal condition of operation for a single valve, and indicates that the plate current is not cut off for any portion of the cycle*“. And * points to a broader definition in Sect. 5(i)B where it is stated “A Class A amplifier is an amplifier in which the grid bias and alternate grid voltages are such that the plate current of the output valves flows at all times”. Therefore if we want to get really strict about this issue we have to remember also that any amplifier with its devices working in Class A can only be considered so if we guarantee that its input will stay within the limits set by the biasing conditions. We can make a single ended amplifier stop working in class A not only toying with the bias but also just using a higher than specified input voltage even with the “class A” bias. Whenever we write we implicitly make some assumptions about the level of knowledge and/or tolerance that the reader may have. It is very hard to take in account what everybody knows or is willing or not to fill in as obvious lapses that could be understood taking in account the context. What is assumed in my statement is that to be used in audio there must be a way to have conduction during the whole 360 degrees of a cycle but it may happen through one device or by making the 360 degree conduction through more than one device. It is the same assumption made in the Radiotron Designer’s Handbook’s first quote.

Comments