It seems that things are more complicated than they look. In question is the Indel 220V / 16V, 20VA transformer; 4.48H / 222Ohm, 102.5mH / 2.7Ohm. I did numerous tests with various frequencies between 25Hz and several thousand Hz with the Sch1 from attachment scheme. At frequencies below a few hundred Hz it begins to become nonlinear below 30mA DC. At frequencies above 1000Hz it is linear right over 70mA DC. I noticed that at a few tens of Hz it starts to become nonlinear below 10mA DC. With my 30V/5A lab. source I can't go beyond 70-80mA in this set up. All currents were read from the source ammeter so they are somehow average values. The duty cycle matters a little, but the frequency strong determines the maximum current for linearity. I would have thought that at low frequencies it saturates at high currents, but it's exactly the opposite.
I have attached the oscillograms for SCH 1 taken on the 4.7Ohm resistor for 242, 470 and 3250Hz so that no limitation or nonlinearity occurs. Voltage values are on the top left screen.
For a better check I thought of a test that is closer to my goal, namely the behavior in the simultaneous presence of a sound signal and a direct current . I used the Sch 2 scheme (see attachment) together with an online sine generator or even music. In principle, with music, when a current of more than 30-40mA DC is introduced, only the low frequency (bass) level decreases. However, the sound does not become distorted at least for the ear. It's not the same when I turn on the oscilloscope. I did tests with a moderate signal, around 1.4V true RMS on a 16Ohm speaker at various frequencies, all sinusoidal between 20 and 20000Hz. Each time I watched to see what happens when I introduce a voltage between 0 and 30V in the two primers transformer connected in series, which corresponds to a current up to 80mA. At low frequencies there is a strong nonlinearity and after 1000-2000Hz the signal remains pure sinus, even at 20000Hz. At a few tens or hundreds of Hz with the increase of the DC current, artefacts appear on both even and odd harmonics. For rectangular input signals, the odd components remain unchanged as the current increases, but even components seem to increase as the DC component increases. For a sinusoidal signal at 500Hz I can introduce a maximum of 30mA DC and from 1000Hz and up I can feed with 70-80mA DC and the wave stays clean without any harmonics. It is possible that there is a connection between these currents and the fact that the transformer data shows a maximum current of 90mA but for 50Hz.
In conclusion, why saturate the toroidal core rather at low frequencies?