WA1GFZ, I think you are saying that you can always pick up a feedback signal with a third winding and that this indicates unbalance and/or the actions of the loss monster. I don't agree that this necessarily proves unbalance or loss monster. There is a third mechanism at work that will always make it possible to pick up a feedback signal with a third winding even if everything is perfectly balanced and the loss monster is asleep. I had not thought of this till just now thinking about your post above. Here's what's going on. As you said, the on transistor is essentially placing a portion of the power supply directly across the upper winding which is acting like the primary of a transformer. This induces a voltage in the secondary winding which is acting like the secondary of a 1:1 transformer. I think we agree so far. Now whatever impedance the lower coil is looking into is directly reflected as the same impedance in the upper coil because of the 1:1 turns ratio of the two coils. Here's why I think you can put a third winding in the transformer to pick up a signal and use it for feedback and it doesn't necessarily mean there is imbalance or the loss monster is responsible. Here's my assessment of why you can do this. I'm betting that you know about the magnetizing current that exists only in the primary winding of a power transformer (actually all transformers), but like me, never applied that thinking to an RF transformer although it does apply and you have probably designed RF transformers taking this into account but may not have realized that you were. Anyway, in any and all transformers that are acting like transformers the primary winding has a magnetizing current that the secondary does not have. Take the case of a 1:1 isolation power transformer. With the secondary open circuited, the primary still has a current flowing. This current is partially due to leakage inductance and partly due to a magnetizing current that has to flow in order to produce a flux so that there is a secondary voltage even when open circuited. Ignore the leakage component on the primary for now. When we first apply power to the primary, a current begins to flow as it would in any inductor. At equilibrium, the applied voltage and the flux always have to satisfy Faraday's law that the applied voltage always equals the time rate of change of the flux times the number of turns in the primary. The current that it takes to set up this magnetizing flux is the magnetizing current and by design is usually a small fraction of the rated load current of the transformer. Now when we apply a load to the secondary winding, a current flows in the secondary and Lenz's law says the current will produce a flux that opposes the flux that generated the voltage in the secondary thus starting to cancel the flux caused by the primary magnetizing current. At this point you can look at what happens several ways. I like to over simplify it a bit and say that since the secondary current tends to cancel or reduce the flux in the core, Faraday's law is no longer satisfied on the primary winding and the primary current begins to increase to bring the flux back up to a value that satisfies Faraday's law for the applied voltage. So now the primary winding has two current components (actually more if we look at all effects). One is the original magnetizing current that is producing the flux in the core that produces the voltage in the secondary, and the other is the primary load current that results from the secondary load current. The net result is that when we connect a load to the secondary, the flux in the core pretty much ends up right where it started except for some relatively minor differences due to leakage and resistance, etc. As the secondary current increases or decreases due to load changes, the primary current increases or decreases in response always keeping the flux in the core essentially constant at whatever value satisfies Faraday's law on the primary winding. The point is, even in a high power transformer, the flux produced by the secondary current is cancelling the flux produced by the primary LOAD CURRENT component and the flux in the core is essentially the same as when the secondary is open circuited. IE; for all practical purposes, the flux in the core does not change when going from no load to full load. If you were to calculate the flux produced by either the secondary or primary winding by looking at the current and the number of turns, you would come up with an enormous flux possibly as much as ten times the saturation level of the core. This is because the flux due to the LOAD current components of the primary and secondary windings ACTUALLY are cancelling one another and what remains is the original magnetizing flux. This never goes away no matter how well balanced everything is. My guess is that you probably know all this and/or have seen equivalent models for real transformers including the magnetizing currents and such. This all applies to RF transformers as well. Classic design criteria for RF transformers say to establish the number of primary turns based on the criteria of making the inductive reactance of the primary winding at least 4 or 5 times the load impedance expected to be reflected into the primary winding at the lowest frequency of interest. What this is doing is setting the magnetizing current of the primary to be 1/4th to 1/5th of the actual signal current of the primary winding. It can be made a smaller fraction of the load current by adding more turns. In the case of power transformers, the magnetizing current can be made to be a very small fraction of the load current. In the case of RF transformers, adding turns adds resistance very fast because of the skin effect so you reach diminishing returns quickly. In the case of a power transformer, we don't change the primary voltage so the magnetizing current and flux is constant and does not change with load. However, in the bifilar feed coil in a linear amp, we do change the amplitude of the applied voltage according to the amplitude of the signal so the magnetizing current and resultant magnetizing flux in an RF transformer WILL be proportional to the RF signal strength even if everything is perfectly balanced. What's the point of all this? We are alternately placing the power supply across one coil then the other. Even if the load currents are perfectly balanced and synched, there is always always always a small magnetizing current and flux in the primary that is never never never completely cancelled by the secondary current even when everything is perfectly balanced. If you follow typical RF transformer design guidelines, this current and resultant flux can be as much as 1/4th to 1/5th of the actual load currents. So you will always be able to pick up a signal with a third coil even with perfect balance of load currents, identical transistors, although this may not be the only reason you are able to pick up the feedback signal with a third winding.