[maxhifi]

To find out the transformer ratio, apply about 400Hz at the amplifier input, and measure the AC voltages at the primary, secondary, and feedback windings of the good transformer with a VTVM.

I like Electonic M's idea of using the speaker voice coil winding for feedback. No significant DC will make it through the tone control so no need for a blocking cap. If it was off the plate that would be another story.

Anyway once you know the transformer ratio, you will know if you need to change anything to use the speaker winding. If it's significantly different, you can change the value of the 680 ohm reaistor up in value to increase the amount of feedback, and down to reduce it. For example if the feedback winding puts out more voltage than the speaker voice coil winding, you would want to increase the value of the 680 ohm reaistor, and vice
Versa.

In a perfect world this should be accompanied by a change in the 3.3k cathode bias reaistor to keep overall current the same.

If you hook it up and it oscillates, swap the voice coil winding leads. Also make sure the voice coil windings are not grounded anywhere.

I wonder if this chassis was also used in a model with one mono woofer and two separate mid/tweeters. If it was that would explain why the feedback winding was separate in the first place. This thing doesn't look less cheap after analyzing the circuit, but I am happy to see it used negative feedback! (That ought to help control the response of those speakers a bit)