Nonlinearities in Vacuum Tubes, Bipolar and, Field-Effect Transistors(2024)

The article is good.

Nonetheless, while several pieces of audio folklore are rightly criticized, one is reproduced with little comment.

> And the bipolar transistor configured as a current gain stage is completely linear, at least as far as modeling goes.

"As far as modeling goes" is right only about extremely inaccurate bipolar transistor models, which are suitable only for mental calculations, not for computer simulations.

A bipolar transistor is a linear current amplifier, where the gain is the so-called beta, only within a narrow range of collector currents. Both at high collector currents and at low collector currents, the value of the current gain decreases quickly, instead of being constant.

For high-power audio amplifiers, this is especially important for the final output transistors. If they are bipolar transistors, they are almost always used in the current range where the current gain drops quickly, so they are non-linear as current amplifiers.

In general, it is far more accurate to consider a bipolar transistor as a transconductance amplifier, where the output current is controlled by the input voltage, exactly like in a field-effect transistor.

There are 2 differences between a bipolar transistor (BJT) and a field-effect transistor (FET), as transconductance amplifiers. The first is that a FET is only mildly non-linear, the transfer function being a polynomial of low order, e.g. of 2nd order, while a BJT is strongly non-linear, the transfer function being exponential. While a BJT is strongly non-linear, it has a much higher transconductance gain than a FET, which is its main advantage in analog applications, so it can be easier linearized by negative feedback.

The second difference is that while a FET has a high input impedance as it is desirable for a transconductance amplifier, a BJT has a current leakage at its input, with a current source that is equal to the collector current divided by beta.

Thinking about a BJT in this way is more efficient than thinking about it as a current amplifier, because this description is valid in a much greater range of collector currents, spanning many orders of magnitude. When the input capacitance, which stores the control voltage, is also taken into account, this description is also valid in a greater range of signal frequencies.

As TFA says, in the series vacuum tubes => FETs => BJTs, the non-linearity of the gain increases. Nonetheless, in the same direction the transconductance gain also increases even more. The higher gain makes easier the compensation of the non-linearities by negative feedback, thus for complete amplifiers, it is easier to obtain lower distortions with the more non-linear devices that have higher gain.

For audio amplifiers, these characteristics are mostly of historical interest, because now one can make switching amplifiers for audio, where the non-linearities of the transistors are irrelevant. The various kinds of distortions in switching amplifiers have other causes than the non-linearities of the power transistors, e.g. they may be caused by slow switching, by non-linearities in integrators, etc.