Are the ways that we’ve come to characterize and name certain types of feedback accurate?
I’ve been bothered for a while about some of the ways that we’ve come to characterize and name certain types of feedback. This is not just an idle rumination. I know of at least one public “flame war” that resulted from what I believe to be faulty inferences drawn from some of these characterizations.
Well, what is feedback? One answer might be that it is a process which senses a signal it wishes to influence and feeds a portion of it back to an earlier point in a circuit where it can exert some form of control. Consider in Figure 1 the four classic arrangements for routing two kinds of signals: feedback and excitation of the amplifier. We say that the feedback source is either shunt derived (voltage, across the load) or series derived (current, through the load, appearing as a voltage across an impedance in series with the load).
We also speak of series and shunt applied feedback, where that signal is in series or shunt with the excitation signal. In the “shunt” case, the two signals meet at the inverting input and the non-inverting input is grounded. In the series case, excitation is applied to the non-inverting input and feedback to the inverting one. Note the ambiguous way that I have drawn and designated the excitation signal sources Sp and Sm. This is on purpose–they are meant to be non-ideal. By this I mean that they can be thought of as either ideal voltage sources in series with impedances or ideal current sources in parallel with ones.
Now consider Figure 2. With suitable choices of values of impedances Zt and Zb and signal levels of sources Sp and Sm, we can implement any of the four circuits of Figure 1. Let’s work with the more versatile circuit in Figure 2.
Figure 2 A more versatile circuit with which to work
The first case we’ll look at is the series applied one (the derived type is irrelevant for our purposes). Source Sp has a non-zero output, and Sm a zero one. As such, Sm is just a connection through its intrinsic impedance Zm to ground. The amplifier output sends a current through the Rf −Rg−Zm network. Suppose we use a real-world op amp like the venerable TL072, one which comes very close to accepting no current into its inputs. As with all op amps, the signal output of its input stage is a current which ultimately controls its output voltage. In this case, that current is sourced from within the op amp itself and is controlled by the difference between the two (excitation and feedback) input voltages. This is clearly voltage feedback, as is traditionally claimed; a voltage signal is fed back to the inverting input where it exerts control. And at this point of control, negligible current enters the amplifier.
Now let’s investigate shunt applied feedback. We’ll just reverse the operations of the sources; Sm now has a non-zero output and Sp is zeroed. Recall that it is claimed that shunt applied is current feedback. Now I ask you: does it make sense that we have changed the type of feedback in a circuit simply by changing the amplitudes of signal sources? Suppose both sources had non-zero outputs – do we then have voltage and current feedback? Suppose both levels were zero – do we then have no feedback? (Surely the op amp output resting near zero volts would object to that conclusion!)
[Continue reading on EDN US: Is this current feedback?]
Christopher Paul has worked in various engineering positions in the communications industry for over 40 years.