How to describe and address certain types of feedback has always been a problem for me. It’s not just a boring fantasy. I know that at least one open “war” was triggered by the wrong inference drawn from these feedback characteristics.

So, what is feedback? One answer is that it refers to a process, that is, to detect a certain signal that you want to influence, and feedback part of it to a certain point in the front of the circuit, so that you can exert some kind of control. Figure 1 shows four classic circuits for two signal routes: feedback and excitation of the amplifier. We say that the feedback source is either derived in parallel (the voltage at both ends of the load) or in series (the current through the load is expressed as the voltage at both ends of the impedance in series with the load).

We also talk about series and parallel feedback, where the signal is in series or parallel with the excitation signal. In the “parallel” case, the two signals converge at the inverting input and the non inverting input is grounded. In the “series” case, the excitation is applied to the non inverting input and fed back to the inverting input. Please note that the way I draw and specify the excitation signal sources SP and SM is a little fuzzy, which I do on purpose because they are not ideal. I mean, they can be considered as ideal voltage sources in series with the impedance or ideal current sources in parallel with the impedance.

Figure 1: four classic circuits for two signal routes.

Now look at Figure 2. By choosing proper impedance values ZT and ZB and source signal levels SP and Sm, we can realize any of the four circuits in Figure 1. Now let’s use the more general circuit in Figure 2.

Figure 2: using a more general circuit.

The first case we want to look at is the concatenated application (independent of the derived type). The source SP has a non-zero output, while SM has a zero output. Therefore, SM is only a connection to ground through its inherent impedance ZM. The amplifier output sends current through RF RG ZM network. Real operational amplifiers, such as the old tl072, can accept almost zero input current. Like all operational amplifiers, the signal output of its input stage is the current that ultimately controls its output voltage. In this case, the current comes from the operational amplifier itself and is controlled by the voltage difference between the two (excitation and feedback) input voltages. Traditionally, this is obviously voltage feedback, where a voltage signal is fed back to the inverting input and controlled there. At this control point, the current into the amplifier is negligible.

Now let’s look at parallel feedback. Let’s flip the signal source, that is, SM now has a non-zero output and SP is reset to zero. Recall that parallel applications are called current feedback. Now I ask you: can we change the feedback type in the circuit by changing the amplitude of the signal source? Assuming that both sources have nonzero outputs, do we still have voltage and current feedback? Suppose both are zero, do we have any feedback（ Of course, if the output of the operational amplifier is stable at about 0V, this conclusion will be strongly countered!)

Is this current feedback?

We continue. This current feedback (?) Put the amplifier in (the incoming current can be ignored, right?) Bypass and terminate at source SM. If the ratio of network impedance to SM impedance is very large, then any effect of it can be ignored. Is this still current feedback? I don’t think so. On the contrary, it is a voltage feedback, and the input voltage of its operational amplifier is now in the state of zero level and close to zero level, but as long as the level of SM is close to the level of SP, their voltage difference is basically the same (which is very important for the operational amplifier).

If this is not convincing enough, the impedance of RF, RG and SM can be doubled. We’ve just halved the current flowing through these devices. Therefore, the assumed current feedback must also be halved. However, the output of the circuit has not changed substantially. Therefore, it is not affected by current feedback.

So, does current feedback really happen? of course. We use a different operational amplifier instead of tl082. The output (signal current) of its input stage comes from a point outside the operational amplifier through its inverse input. For the feedback part of the signal, this point is the output of the operational amplifier. Almost any device that connects one or more transmitters to this input meets the requirements (ssm2019 is one option, and the current feedback amplifier known in the industry as CFA is another). Almost all the devices entering the transmitter will exit its associated collector at a certain point in the operational amplifier, which constitutes the output of the input stage and determines the output of the amplifier. With a parallel application configuration, the amplifier still drives current through the network and terminates at the source SM (in most cases). But this time, a small part of the current is “stripped” and fed back to the inverting input of the operational amplifier, thus exerting the required control over the output of the device. Of course, this is current feedback because it meets the consistent requirements of parallel applications.

Finally, back to the case of series application, the operation of the operational amplifier is of course unchanged. In fact, the output signal current of the input stage still comes from the outside of the amplifier – only from the output of the operational amplifier. In order to see more clearly, replace the circuit seen at the inverting input with Thevenin equivalent circuit: feed the signal source with voltage Vout · RG / (RG + RF) through a resistor with resistance value of RF · RG / (RG + RF). I think it’s still current feedback – the output current of the amplifier’s input stage comes from the decaying output of the op amp through a single resistor.

You may argue that the ssm2019 / CFA amplifier is still sensitive to the differential input voltage, so it is still a voltage feedback device (the only impossible way is that if its low inverting input impedance is zero, there will be no input voltage difference in this case). But if we agree with this argument, we must also admit that there is a very large but still limited impedance between the inputs of tl082. This means that a current will be generated between these inputs, so tl082 is a current feedback device!

This reasoning is untenable. Instead, I recommend classifying circuits by feedback type by analyzing the current source from the input stage of the amplifier. If the current source is in the operational amplifier, there is no current feedback from the output, so the operational amplifier and circuit must work with voltage feedback. If the current source comes from the inverting input outside the operational amplifier, we consider it as current feedback.

Due to the above reasons, we should correct the view that “series application is always voltage feedback, parallel application is always current feedback”. Whether it is voltage feedback or current feedback depends on the amplifier of the circuit rather than its topology.

Editor in charge: GT