So, I was led to redo a PCB for an eight wire decoder, the wires of which will have to be soldered individually at the indicated locations. I did not put a connector in order to simplify the circuit design.
New electrical diagram.
The marks of the rail, motor, and LED connections are taken from the original circuit.
New electrical circuit : top, tracks; middle, components; bottom, both together..
Some devices (capacitors, inductors) are to be got back from the old PCB. The others are easily found in e-commerce. Note: marking of SMD resistors: 2.2 kΩ (2k2): 222; 4.7 kΩ (4k7): 472.
Note: on the (vectorial) high definition drawing, the etching mask is shown upside down. Indeed, when exposing the board (thickness 0.8), the mask inking must be directly in contact with the board. If the light had to go through the polyester film thickness, under the ink, it could diffuse and give fuzzy outlines.
I simply glued on the LEDs, with Kristal Klear, a small piece of translucent orange plastic cut into a soft ring binder cover. At first, I wondered if this extra thickness would prevent the body to be correctly reassembled, but no… However, there is still a small problem: LEDs are too punctual, not exactly centred, and so the light intensity is not well-balanced between the two headlights of the same side.
Many of us have noticed the relative slowness of this machine at maximum speed. Mine crawls at 113 km/h to scale under 12 V DC. In DCC, CV5 = 255 (maximum) gives 150 km/h, which is too much, but the speed collapses to 120 km/h as soon as CV5 = 254. Xavier Geillon, on the Loco-Revue forum, suggested setting the EMF divider (CV50 = 32), and it works! Then, by setting CV5 to 186, I get 140 km/h (Lenz Silver decoder).