Even if transistor internal impedance is only tangentially related to the NanoVNA, I would like to weigh in on this matter. Particularly knowing that there are a lot of hams here.
In a power amplifier, a transistor typically goes through two or three very different conditions, in each RF cycle. Disregarding internal capacitances and inductances, which is a good approximation at relatively low frequencies, they are like this:
- While the transistor is not conducting, it has an infinite internal impedance.
- While it is saturated, it has a very low internal resistance, roughly like 0.1? for a BLF188.
- While it is in active conduction, the gate voltage basically controls the drain current, so its behaviour is close to that of a current source, but not fully. Consequently its internal resistance at this time is pretty high, but not anywhere near infinity, and varies according to the instantaneous gate and drain voltages.
If the amplifier circuit has negative feedback implemented by taking a sample of the drain voltage and injecting a proportional current into the gate drive circuit, as is often done, then this tends to make the amplifier control the drain voltage as a linear function of the drive voltage. In this case the internal resistance of the transistor+feedback cell during active conduction is very much lower than in an amplifier free from this kind of feedback. If this feedback is strong enough, it can turn the transistor+feedback combination into a pretty stiff voltage source, that is, give it a very low internal resistance.
Another way to implement negative feedback is by inserting a small resistance in the source lead (or emitter, but BJTs are pretty obsolete now...). In fact all practical transistors have some such resistance built in, inevitably. This kind of negative feedback intensifies the current-source behaviour of a transistor, driving its internal resistance up instead of down, from the no-feedback value.
Regardless of what kind and amount of negative feedback an amplifier has, the impedance of its transistor during saturation or OFF condition are the same as without any feedback, of course. The feedback can act only while the transistor is operating in its active range.
When we make a push-pull amplifier, things get more complex. Because such a push-pull amplifier might have the two drains stiffly coupled together (in opposite phase, of course), by means of a true center-tapped transformer, or a bifiliar feed transformer, or it might have uncoupled drains, typically implemented by using two separate feed chokes, or (often unknowingly) by using a center-tapped transformer that isn't truly working as such.
Since the load of a push-pull amplifier connects in a balanced fashion, between the two drains, the internal impedance of this contraption depends very much on drain-to-drain coupling. Let me put this in a little table, even if the format of the table might be messed up depending on what font you are using to see it:
Transistor 1 Transistor 2 Z with coupled drains Z with uncoupled drains
==============================================================
off off infinite infinite
active off feedback-dependent infinite
saturated off very low infinite
active active feedback-dependent feedback-dependent
active saturated very low feedback-dependent
saturated saturated forbidden! very low
If all this looks too complex, don't worry too much, because anyway we don't need to match the load to the internal resistance of our transistors. In addition to being pointless, it would be impossible! What we need to do is what Fran?ois already wrote: We need to present such a load resistance to our transistors that at the maximum acceptable drain RMS voltage (which depends on supply voltage and the distortion level that we are willing to accept), this load resistance will take the desired power. In the interest of efficiency, we actually want our transistor's internal resistance to be as DIFFERENT as possible from our load resistance! Zero ohm is great, infinite is also great, and that's why class D amplifiers are theoretically 100% efficient: They switch the transistors between saturation and OFF condition.
An important point that many amateur power amplifier designers miss, is that we need to care not only about the load impedance at the operating frequency, but also at the harmonics. Since in almost any real RF power amplifier the transistors spend lots of time OFF, and sometimes some time saturated, and also in active mode they are quite nonlinear, there is always a lot of harmonic contents, which we don't want. So, while our matching network must present the proper load resistance to the transistor at the operating frequency, it needs to present either near-infinite or near-zero impedance at harmonic frequencies, if we want an efficient, clean amplifier. But it does matter a lot which option we choose: Very high or very low impedance! Our selection MUST properly match the characteristics of the amplifier configuration! For mainly high internal impedance, current-source configurations we need a matching circuit that shorts out the harmonics, such as a parallel tank or a Pi filter, while for a mostly low-impedance, voltage-sourcing configuration we need an output network that presents high impedances at the harmonics, such as a series tank or a T network. If we choose the wrong type, the efficiency of the amplifier will be much poorer than expected.
Everything I wrote until here is valid as long as the frequency is low enough to disregard connection and device capacitances and inductances. Starting roughly in the medium HF range, this is no longer the case, and at UHF it definitely isn't! The principles remain the same, of course, but we need to calculate the load resistance that we will make appear right in the silicon chip of our transistor, taking into account its capacitance, the lead inductances, and all external wiring reactances. This tends to get quite complex.
In push-pull amplifiers that operate over a wide bandwidth, such as 1.8-30MHz, an additional difficulty is that the degree of drain-to-drain coupling very often changes dramatically over the frequency range. If the design was made to use coupled drains, almost always the amplifier ends up having coupled drains only on the lower bands, and transiting into the behaviour of an amplifier with uncoupled drains on the higher bands. It might require one type of low pass filters on the low bands and a different type on the high bands, and also different biasing, to get optimal performance on each band.
Also any signal runtime between the drains and the filters or matching circuits causes a different phase shift, and thus impedance transformation, on each harmonic. In that case, for example a Pi-type LP filter (low impedance at harmonics) might appear as any crazy reactance, or even infinite impedance, on some harmonics, totally messing up the amplifier. In HF power amplifiers this becomes a really big problem when using transmission line transformers, because these cause a long delay. Many designers end up using diplexer filters, because they find no other way to solve the problem. But diplexer filters increase the complexity and reduce the efficiency.
Given the variability of a transistor's internal impedance over the RF cycle, it should be obvious that there is no way to measure it with the NanoVNA (or any VNA) directly. It's better to just calculate the required load impedance, design and build the network, then replace the transistor by a resistor of the calculated value, and measure with the NanoVNA into the 50? output of the network, to confirm that it's working as expected. If the amplifier has to operate above roughly 15MHz, it becomes necessary to use an RLC circuit that emulates not only the desired load resistance, but also the internal L and C of the transistor. And it has to be understood that the resistor is NOT an emulation of the transistor's internal resistance, but stands strictly for the load resistance that we want to apply on the transistor!
Manfred
F4WCV
Donald S Brant Jr
Lou W7HV
Andrea I2UEA
