Chapter 7: Tubes vs. Transistors

I have designed both tube and transistor amplifiers and was fortunate enough to be a young man at the time transistors were just coming out. Making both tube and transistor amplifiers allowed me to understand their differences in circuit design, as well as how their various characteristics of distortion and noise differ, and of course, how they sounded.

When we talk about tubes, we need to consider the different kinds of tubes that exist for audio purposes. Diodes were used primarily for rectifiers, triodes were the first amplifying tubes, tetrodes were rarely used, and pentodes were an improvement over tetrodes and actually very popular audio output tubes. Pentodes are also very popular in pre-amplifiers, when a great deal of gain is required. Even though pentodes are somewhat looked down upon, there are many famous tube pre-amps, both commercial and those for the public, that use pentodes, QUAD of England being one of them.

Now, a tube, especially a triode, is a very "linear amplifying device". Meaning that in a proper circuit, the voltage going in is multiplied by a constant number to the voltage going out. What that means is if the number of multiplications of amplification is constant, there will be no distortion. For example, 1V coming in produces 10V going out and 2V going in produces 20V going out. Also, as current in the tube increases, the amplification factor increases slightly, and we get second harmonic distortion that also causes inter-modulation distortion (IMD). In a triode, this distortion can be quite low such that in a typical line stage amplifier putting out 1V, most tubes will only have about 0.1% distortion. This distortion is proportional to output levels so that at 2V, it will be about 0.2% and at 3V, about 0.3%. This is generally low enough distortion that we do not need to use feedback and we can use the device just as it sits in a very, very simple circuit.

On the other hand, a transistor is a current amplifying device. As such, we must make certain circuit changes because we are generally amplifying voltages from sources to other voltages. If we were living in a current world, the transistor would be a more desirable device to use directly. To make transistors linear for voltage amplification, we either must degenerate this current characteristic or we must use a great deal of feedback, the latter of which is generally what is used.

Most transistors also end up in complex circuits. For example, op-amps, that can be made with discreet components. It takes about ten transistors to make an op-amp, or they can be integrated to create a circuit op-amp that may have twenty transistors to get a little higher performance, but generally at the expense of having a great deal of gain, say 120 decibels (dB). If we reduce that gain to 20 dB we would need 100 dB of feedback. That is a lot of feedback although there is nothing wrong with that if it is done properly. Op-amps have another problem in that there are thermal characteristics going on because the chips are so small the input transistors and the output transistors are virtually next to each other. The currents that are changing in the output transistors as the output signal is produced are changing the environment of the input transistors and some very strange modulations may occur. This may be why integrated circuits are not thought to sound as good as discreet circuits with transistors where the input and output transistors can be separate.

One of the things about tube circuits, such as that for a pre-amplifier line stage, is it can be done with one singe triode, a couple of resistors, and one capacitor. You cannot do that with transistors. With a transistor, such as a bi-polar transistor, you need several more components and you need to do something about the fact that the input impedance of this device is rather low, and the output impedance is very high. These are two things we do not want. A tube has virtually infinite input impedance because the grid is virtually an open circuit. The triode has a low output impedance that is known as its plate resistance. The 6922 having one of the lowest plate resistances at about 3,000 ohms, while a 12AU7 is around 10,000 ohms, and a 12AX7 is as high as 100,000 ohms. We really desire low output impedance which is why a tube like the 6922 is popular.

The transistor on the other hand has almost an infinite output impedance and a very low input impedance of around 2,500 ohms. To increase the input impedance, we degenerate the circuit with resistors, or we make circuits with feedback both of which increases the input impedance. These may or may not be good things to do, but either way it always takes quite a few more parts. The one exception is an FET that acts very much like a pentode tube. It will have virtually infinite input impedance but also have virtually infinite output impedance. That becomes a problem. A FET is not an incredibly linear device, and some people just feel that FETs do not sound very good. For that reason, we do not see FETs in line amplifiers very much.

Another advantage of vacuum tubes is that the circuits are so simple that they can be hand wired. The parts are large enough, we have tube sockets, and everything for the tube age of audio was meant to be hand wired. Point to point wiring lasts much longer, is more stable, and will not have stray paths. Transistor circuits can also be hand wired but is rarely done as the complexity of the circuits and the small parts lend themselves to circuit boards. Of course, integrated circuits are almost always on circuit boards because of their small size. These are also lower voltage devices. This brings up another point, a transistor or integrated circuit line amplifier can generally only put out 10V RMS because the power supplies are typically plus or minus 15V. That results in a maximum 10V RMS signal. Whereas tube line amplifiers having hundreds of volts of B+ can often put out 30, 50, and 60V. Not that this is necessary, but it certainly gives us a lot of head room. Also, because the voltage on tubes is so high and we are using such a small fraction of the range, we are in a very linear part of the curve.

What I have said so far applies mainly to pre-amplifiers, but if we look at power amplifiers other considerations occur. One of the greatest problems with solid-state power amplifiers is protecting the output devices from short circuits or very low impedance flows. Transistors are not at all tolerant of having too much current flow through them. The transistor can short in a millisecond. To make successful transistor amplifiers that are short circuit proof either relay protection circuits must be used, or more commonly, low lying limited circuits. In any event output transistors must be protected and are not free from fault if there are short circuits. With vacuum tube amplifiers the tubes can take short term overloads of very high value with no damage whatsoever. As such, vacuum tube amplifiers do not need current limiting devices. This makes them ideal for driving difficult loads, such as electrostatic speakers that not only go to low impedances, but also have reactive phase angles, that often will trip the protectors in a solid-state amplifier. A tube amplifier will just cruise on through that. Although, it may run out of current and soft clip, it will not be as noticeable as a transistor amplifier current limiting. Also, since tubes are more linear than transistors, tube amplifiers require less feedback. A typical tube amplifier can have anywhere from zero feedback up to usually no more than 20 dB. Whereas a transistor amplifier may have 40, 60, 80, and 100 dB of feedback, again to reduce distortion and lower the output impedance of the amplifier.

Tube power amplifiers come in three flavors: Triode, pentode, and a third in between being ultra linear. Triode amplifiers tend to have more distortion, but also have inherently low output impedance because the triode itself has inherently low output impedance. These can be made single ended or push pull. Pentode amplifiers have high impedance, which means they will need a little bit of feedback. Ultra linear amplifiers are somewhat in between, getting some feedback from their screen grids. Transistor amplifiers also come in two flavors. One being bi-polar, which were the original transistors, and later the FET transistors, that have very different characteristics. It is very difficult to make FET amplifiers with stable bias and this has been a problem. There are also two kinds of FETs, the linear ones Hitachi made and that were used in the early Halfler amplifiers, then the switching types that are not at all linear and are used in most modern amplifiers. These are very hard to stabilize the bias and because of their very rapid turn on voltage, these devices may not sound as good as bi-polar devices.

One of the things you might look for in buying a solid-state amplifier is a multiplicity of output transistors. It turns out that the more output transistors you have the higher current will be available, and perhaps if the current is high enough the transistor can be protected with fuses and there will not be any current limiting. This is, I believe, the best way to make a transistor amplifier, with no current limiting, and enough transistors so that the fuse alone can protect them. Tube amplifiers can also be made with a multiplicity of output tubes, but this is generally for a different reason, either to run the bias higher up to Class A, or to provide more current. Again, short circuit limiting is generally not an issue. I have not seen any vacuum tube power amplifiers that have short circuit limiting because they just do not need it. It is interesting to note that low noise tubes can have virtually as low a noise, and sometimes lower noise levels than FETs or transistors in many circuits.

One advantage of transistors is they will generally stay low noise, and with vacuum tubes they may have to be replaced when they get noisy. Vacuum tubes also have microphonic characteristics that are not desirable but by selection we can choose tubes that are both low in noise and low in microphonics. A low noise tube, like a low noise FET, must have very high transconductance. This is one measure of the gain of the tube. The higher the transconductance the lower the noise. One thing that is a misconception with matching of vacuum tubes is many people are matching for transconductance because this is what the popular Hickok tube testers measure. Simpler tube testers merely measure emission and while that is suitable for knowing if the tube is good or bad, it does not tell the whole story.

Transconductance testers, although people think they are far more sophisticated, are not necessarily so. Matching the transconductance of two tubes does not in any way guarantee that their voltage gain will be the same. This is especially true for triodes. In a triode the gain is called Mu which is simply its voltage amplification, and it is the product of the transconductance along with the plate resistance. Measuring those two alone will not guarantee equal Mu of the tube. We at RAM Tubes and others who know better will measure the actual voltage gain of the tube and report that. It is the voltage gain of the tubes that you would want to match to have the tubes of your amplifier be of equal gain.

One must keep in mind that the important measurement of a triode is the Mu. It is also interesting to note that with most manufacturers the Mu of a triode is usually graded within about a plus or minus 10% range. On the other hand, transconductance, or GM, is the important measurement of a pentode. Pentodes do not have Mu so it is only the transconductance that matters. Keep in mind the triode is a low impedance device. A pentode is a high impedance device. In triode circuits the plate resistance along with the transconductance determines the gain. In pentode circuits it is the transconductance along with the plate load resistor that determines the gain.

With transistors, FETs are much like pentodes. Bi-polar transistors are unlike any of these other devices. It has a current gain which is called Beta. Even more interesting, Beta varies a great deal from one transistor to another, even if they may come from the same batch. Whereas a tubes transconductance or Mu can be generally controlled within 10%. A transistors current gain is rarely controlled better than 300%. You could easily have transistors whose gain could be as low as 100 and as high as 300 coming out of the same batch. This is simply because the Beta of a transistor is very difficult to control and in circuits where the Beta is important the Beta is generally matched. The Japanese have done this in a lot of their amplifiers, and you will see little colored dots on top of the transistors such as red, green, or yellow, telling you which Beta range it belongs to. These colors were painted on, measured by hand, and binned out into different Beta bins. FETs have a similar large variation in what is called the "gate pinch off voltage". This can vary as much as 300% making the matching of FETs very difficult, especially in power amplifiers. If you put a bunch of FETs in parallel it is very difficult to get them all to conduct the same amount of current. It is much easier with bi-polar transistors to get them to share current equally. These are all problems that are left up to the circuit designer and the resulting amplifier will have a lot to do with the ability to manage these variations.