Do you have some references regarding the benefits you claim of the stored electromagnetic energy in the core for peaks? In general, you do not want DC flowing through a transformer, it may require a gapped core in order to avoid saturation. And almost all gapped-core transformers exhibit inferior specs when compared to their non-gapped counterpart. Also, consider that the way current flows and the way the coils are wound around the core may cancel this so-called stored EM energy in the core, so this EM energy in the core might not even be applicable to the configuration you mention, since current in the windings is flowing in opposite directions.
Adding DC to a transformer (which I recognize it might not necessarily apply to a push-pull stage due to its symmetry) adds an offset to the hysteresis curve, which will produce asymmetries that, in turn, introduce even-order harmonics.
From the Handbook for Sound Engineers: https://www.jensen-transformers.com/wp- ... hapter.pdf
Bottomline: do not add DC to a transformer unless strictly required.If dc current flows in a winding, the operating point will shift to a point on the loop away from the center. This causes the distortion of a superimposed ac signal to become non-symmetrical. Non-symmetrical distortions produce even-order harmonics such as second and fourth. When a small dc current flows in a winding, under say 1% of the saturation value, the effect is to add even-order harmonics to the normal odd-order content of the hysteresis distortion, which affects mostly low-level signals. The same effects occur when the core becomes weakly magnetized, as could happen via the brief accidental application of dc to a winding for example. However, the narrow B-H loop indicates that only a weak residual field would remain even if a magnetizing force strong enough to saturate the core were applied and then removed.
NFB does roll-off with frequency, but as long as you have a good amount of gain this shouldn't be a problem. If your unity-gain frequency is high enough, NFB should do its job. Also, the distortion levels with NFB are usually at least an order of magnitude lower than without NFB, so even if you get a more abrupt clipping behavior, you still get far less distortion the rest of the time.
What about the tons of NFB in the analog stages inside your ADC?
I disagree. NFB does cease to be effective if you saturate, but the results are far superior to not using it. If you use transistors without negative feedback distortion becomes much more dependent on signal level as you abandon the small-signal region. By the time you are near clipping, you already have a good amount of distortion produced by the transistors themselves added to transformer distortion. So it is really irrelevant whether NFB clips more drastically when you are introducing distortion most of the time by not using it.Tubetec wrote: ↑Thu Jan 02, 2025 7:18 am Theres also the simplicity and the fact that you can swing four times the supply voltage across the output devices in push pull .
As we said complimentary pairs are fundamentally different devices , and usually a large amount of overall negative feedback is needed to iron things out , when an amp like that starts to overload the NFB can longer work as intended , everything goes hellter skelter .
Another downside of the class B push pull arrangement is the transistors are likely to need matching to some degree , or performance could suffer .
Perhaps its less of an issue with class a bias , as the transistors are always on .
Also, you do not get away with having to iron things out with a transformer. The cancellation of harmonics in a push-pull stage is predicated on the fact that the devices are the same. Transistors have a wide tolerance, threshold voltages can vary wildly, and the large signal characteristics may also differ, so you are still left with the problem of having two different transistors but without NFB to help you out.
The THD levels that can be accomplished with a transformless push-pull stage are several orders of magnitude lower than those accomplished using transformers and no NFB. No one really uses class B as you describe, it is usually some sort of light Class AB. Self suggests biasing the transistors while injecting a signal and measuring THD. The correct bias point is that which minimizes THD at the minimum possible bias current. Note that you also get more distortion when biasing deeper into Class AB due to the transition of the transistors between Class A and Class B mode of operation. That is, when the transistors turn on and off. I see no reason why your arrangement would be inmune to this effect.
Honestly, I see no other reason to abandon NFB and adding a transformer for the application you are referring to other than subjectivity. That is not to say that I am against transformers. I believe that in some specific applications they might be great and necessary, but not in the one you describe.
