So, I was trying to optimize the amount of transformer windings:
For the battery to be usable at lower voltages, the cores need more turns. However it is more work, so there must be a point of diminishing returns as battery life isn't a super important goal in this project.
I was winding one or two transformers with 20, 25, 30 turns, and testing them. If they looked promissing, I would wind one or two more to complete a byte.
However, when winding the 3rd 30 turn transformer, something threw me off. The third one had quite a different waveform. I thought maybe I took a toroid from another batch?
Took me 18 hours to find out, but the glue I was using to glue the toroid on the 3D print changes the characteristics of the transformers.. Facepalm. I had to let the glue cure before measuring them.