Design Considerations: Why tubes sound better
by James Bongiorno Great American Sound Co.
First of all, I would like everyone to know that I’m writing from a position of subdued authority, this authority of course is vested in me by myself. Now for all of you out there in audio city who want to scream egomaniac, do it. Let us, therefore pass on to more important matters, that is, facts. Facts are always there, bold yet mysterious, and though we can acknowledge the awareness of the fact’s presence, we cannot always specifically define what it is exactly.
The first fact is that the human ear is the most incredible piece of instrumentation that we have in aural science. No amount of electronic instrumentation has yet achieved an awareness like our own ears can perceive. Obviously the ear is not a quantitative instrument but rather a qualitative instrument. One of its greatest abilities is in perception and memory of ratios. A ratio is a measure of difference, and that is just what the ear perceives – difference.
Now for the first tomato in my act. Practically all engineers that I’ve ever known fail miserably in at least one aspect of their endeavors; that is interpretation. But then again, maybe they don’t know or are not aware of what they’re supposed to interpret. Anything that the ear can perceive can be measured or evaluated if proper interpretation is sought. Now we are at the point of discussing: Is there a difference between tubes and transistors?
The answer is of course yes. I’m not going to try to tell you that tubes are better, because in some areas they definitely fall flat on the floor – areas such as longevity, changes in characteristics with age, heat dissipation, physical size and bulkiness, etc. But what about the sound? The answer is not „Yes, they sound better“, but rather, „The very best tube equipment sounds better“.
I haven’t designed a tube amplifier in more than 40 years (since I was 17 and that wasn’t very good) and I’m still trying to design better solid state amplifiers (not better than the tube amp I built when I was 17, but just plain better). Before you jump to any rash conclusions (such as, „all solid-state amplifiers are lousy“) I must say that this is simply not true. Solid-state amplifiers are exceptionally good and they’re getting better because some of us are learning. It takes a lot of years of learning.
I will go a step further and state (of course it’s only my professional opinion) that the very best solid-state amplifiers available today are superior in just about every way to what is available in tube amplifiers. There is, however, one area where tubes still hold a lead and that is the ability to drive reactive loads without damage or limiting.
I know of only one solid state amp, one which I designed for another company that has the capability of driving any phase angle from 0 to 90 degrees. And that capability costs a lot of moola .. with tubes, you get it free. This unit however, still does not possess the full capability’that I would like to see.
Why is this capability (0 to 90 degrees) needed and why is it so important?
All loudspeakers are reactive. They are motors and, as such, they cause havoc with many, if not most, solid-state amplifiers. You would be very surprised indeed, to find out just how much the protection circuits are being activated in most solid state amplifiers. The worst part of this problem is the fact that is causes bad sound. It is also hazardous to the amplifier.
It is very sad that most engineers seem to think that the amplifier is going to be driving an 8 ohm load resistor. When I design and make an amplifier, I measure it while driving 100 percent pure capacitive and 100 percent pure inductive loads. I also measure it while driving all power levels, up to clipping, into a loudspeaker load. These are tests that are most important, and resistive loads are used merely, for indicating nominal power output and distortion – that’s all.
I have mentioned the very best tube amplifier available. I am of course referring to a unit that I custom made and is not commercially available. At this point I would also like to stress that I will not supply the plans or schematics for this critter because there are no available transformers. We use it merely as a reference unit in our work. As a matter of fact, a couple of Marantz 9s are just about as good and, if tweaked up, they’re hard to beat. My reference tube unit has a bottom end that just reaches down to the center of the earth – it is the most solid bass that I’ve ever heard. The mids and highs are so silky smooth that it makes me cry. I am going to shock you further. This tube unit has no negative feedback around the transformer – uncanny. A hah – negative feedback, maybe like women and money, the root of all evil? Not necessarily so, but negative feedback can be an enemy if not carefully used.
Tubes and transistors are completely different animals. And we are at a decided disadvantage with transistors. Tubes are natural high frequency devices which transistors are not. Tubes do their thing naturally where transistors have to be made to do things that are not necessairily right for them. Tubes, for the most part, have even harmonic distortion products while transistors, being exponential in nature, have an odd harmonic nature. Transistors have a transit delay phenomenon which is due to interjunction capacitance and if operating points are not carefully chosen, the sound quality can be just awful. Tubes on the other hand, do not suffer this problem because by nature, they are depletion devices. To sum up the problems as I see them, the three areas of hardship with transistors involve:
1. Higher order odd harmonic generation,
2. Transit delay (not to be confused with phase shift),
3. Reactive load limiting.
Tubes are basically square-law devices and as such they develop mostly even harmonics. That is not to say that they don’t have any odd harmonics, because they do. However, it is much easier to find and adjust operating points to balance out these odd-order harmonics. It is extremely difficult on the other hand to eliminate odd harmonics from transistors because they are exponential devices and they produce odd harmonics naturally and they do it over most of their operating range – which is tens of decades of current. If these elements are developed in the early stage of any amplifier, such as the input stage itself, it is very hard to rely on feedback to lower the distortion content because these odd harmonics can and do become the input signal itself. Secondly, the circuit topography can have a lot to do with odd harmonic generation. Thirdly, when we think of tubes, we think of large voltage excursions with relatively small ratios of current changes. In transistor circuits however, the voltage excursions are generally much smaller, and the ratio of changes in current are extremely large as compared with tubes. Since transistors are current operated devices, we are „looking through“ forward-biased junctions (exponential) at all times where, with tubes, we are looking into a reserved biased grid – strictly depletion mode operation. The answer then to this first problem will be fo find ways of lowering the amounts of higher order odd harmonic distortion produced directly by the transistors.
The problem of transit delay is probably the least of the three aforementioned evils and is also the easiest to cure. Even though none of my designs exhibit this problem, it is very surprising to see that most of the power amps presently available suffer, in varying degrees, this headache. Transit delay manifests itself as a sort of masking of definition. In other words, it is like looking through a dirty window rather than a clean one. The problem is also greatly magnified by the use of large negative feedback factors. When an impulse is presented to an amplifying stage, it cannot respond instantaneously since the junction capacitances appear as a short circuit and therefore there is a slight time delay before the transistor can respond. In a tube stage we may be dealing with a few picofarads; in a transistor circuit, we are dealing with hundreds of picofarads. (This of course, depends on the individual transistor). Most triple diffused devices suffer from the „Early effect“, which is aggravated even more by poor designs that have not been compensated for transit delay. This effect is not to be confused with the phase shift present in tube amplifiers that is caused by the passive elements such as the output transformer. In transistor circuitry the problem is caused directly by the active devices themselves which, of course, have gain, and this amplifies the problem more. Again, having proper operating points and circuit topography are of prime importance in eliminating this problem.
The last problem, which is the inability of practically all solid-state amplifiers to drive reactive loads, is probably the worst one and the most offensive one is not to be construed as cross-over notch distortion, which is in my opinion, nowhere near as „unsonic“ as reactive load limiting. If I were to draw a rectangular plot of voltage versus current, positive and negative, we would find that a single tube in an output stage would be perfectly happy operating in all four quadrants. A transistor, on the other hand, gets very unhappy if forced to operate in other than the two positive quadrants and in some cases, it doesn’t even like that. Tubes have a safe area while transistors cry for help at only fractions of this abuse. One might consider, for example, an electrostatic loudspeaker which in some cases (I know of at least two) looks like around 20 mfd load. At 20,000 Hz this is equivalent to 0.4 ohms reactive lagging.
Practically all amplifiers, even most tube units, will have a hard time driving this load, but a tube unit doesn’t have any limiters activating when an impulse occurs. One must understand that if the limiters in a high-power amplifier were removed, fuse blowing or destruction would be commonplace. This is solely because of the fact that transistors have no safe area operating capability compared to tubes. In other words, it would probably take three to five times more output transistors than are already there in order to equal a mere four output tubes.
We are saved by the bell however, the number of loudspeakers around which present this horrifying a load are few and far between.
I’m not going to say much about preamps at the moment because I’m still working on ours. I will say that the same problems exist in transistor preamps as with power amps, however, the manifestations are different. I will say that for some strange reason, which I have not yet discovered, mediocre solid-state power amps are tolerable where mediocre solid-state preamps are not. I have not yet heard one that I like completely, but as with power amplifiers, we’re getting much closer. Anyway, as far as preamps are concerned, I don’t think you will have to hold your breath much longer.