See Figure 1. Original wiring of the Märklin 37645.

I measured the serial resistors for all the LEDs. In the short end of the loco (to the left in the figure above), the resistors are exposed and the values can be read on them, but I double-checked with my multi-meter anyway. I also used the multi-meter to find out the schematics of the LED boards.

Initial testing of the LokPilot
Before I connected the LokPilot to the loco, I verified that I could communicate with it. To do this, I connected two LEDs with series resistors to the two light outputs of the LokPilot.

Figure 2. LEDs connected to LokPilot for initial testing.

At first, I could not get the LokPilot to react on the signals from my control equipment. I tried several times, but the LokPilot seemed to be totally "dead". The yellow LED was lighted, that was all. Finally, I tried with AC analog operation, and then I could get the lights to change. So, the LokPilot lived after all. Trying again with digital commands, I now succeeded to change the light direction and to turn the lights on and off. I experienced this "dead" condition twice more during my tests, but it never happens if I make a proper reset of my control equipment with the track power connected. I think the LokPilot may get confused about the coding if it doesn't see a Märklin/Motorola idle sequence after a long (several minutes at least) power off.

Now when I know I can communicate with the LokPilot, it is time to test programming it. The programming sequence when using a Märklin 6021 control unit is described in the instructions coming with the LokPilot. I use home made equipment that more or less emulates the 6021, and I just had to follow the instructions. It didn't work immediately, however. After consulting the Manual of the New Märklin/Motorola Format by Andrea Scorzoni, I found out that I had implemented the direction change of the 6021 incorrectly. When I fixed this in my control program, I could set the LokPilot to programming mode.

The steps to set the LokPilot into programming mode are the following:

• Put the loco on the track (or in this test set up, connect the LokPilot to the control unit/booster).
• Turn on the power. The LokPilot now recognizes "normal" Märklin/Motorola commands.
• Select the currently configured address (default address is 3).
• Disconnect the power to the loco/LokPilot. With 6021 you press "stop".
• Connect the power while sending a direction change to the LokPilot. With the 6021, you turn the control knob counter-clockwise and press "go". The LokPilot must see either an old Motorola direction change or a mixed mode absolute direction change code. Initially, I only sent the "absolute direction speed 0" codes from my control equipment, except for a hard coded special case to handle the old Motorola format of the original 37645 decoder. The lights should now flash to indicate programming mode. 
• Send speed 0. With 6021, release the control knob.
• Select the parameter number you want to program.
• Send speed 0. Requires no action with the 6021.
• Send direction change. With 6021, turn the control knob counter-clockwise. The lights should now double-flash.
• Send speed 0. With 6021, release the control knob.
• select the parameter value you want to set.
• Send speed 0. No action with 6021.
• Send direction change. With 6021, turn the control knob counter-clockwise. The lights should now shine steadily for about one second, and then revert to flashing.
• Send speed 0. With 6021, release the control knob.

This sequence may seem complicated, but it is actually quite simple once you get used to it. I considered writing a program for my control equipment, that could take all the
steps automatically, but I found it not worth the effort. After all I'm not intending to re-program the loco every day.

Mounting the LokPilot provisionally
I removed the original decoder from the loco, and connected the LokPilot provisionally. I still used the two LEDs described above instead of the original LED boards. While removing the old controller, one of the brown wires from the bogies fell off. It was really difficult to solder it in place again. There must have been a bad soldering from the beginning.

Finding the optimal regulation parameters
This was the really tricky part of the conversion. To test the speed characteristics, I built a test layout on the floor, with 5% grades as well as flat sections. To load the loco, I used four 48010/48012/48020 freight cars, since they are the heaviest wagons I have. Of cause, I tested both with and without load. To start with, I set the recommended parameters for the Märklin BR55/17, because I thought that would be rather close. It wasn't. The first test was disappointing. The loco was jerky and the maximum speed was too low. After a lot of testing, I came to some values that gave acceptable performance, but it was still far from perfect. When i tested the AC performance, and found that the loco ran all too fast, I considered to abandon the conversion and re-install the original decoder. But then I got the idea to test with resistors in series with the motor. I tried some different values, and found that 33 Ohm was a suitable value. This not only cured the AC performance. After going through extensive testing of different parameters again, I reached digital performance that was much better than before. The optimal parameters without the resistor were not good at all with the resistor mounted, so I didn't see the digital improvement immediately.

Calculating the power rating of the series resistor
The resistors I used for testing were standard resistors. I think their power rating is 0.5 W. I suspected that this power rating would not be sufficient for a permanent installation. To calculate the required power rating, I did two things: I measured the DC resistance of the motor: 20 Ohm, and I measured the current required for different load conditions. The maximum current I measured was 270 mA, which gives a 2.4 W power rating for the 33 Ohm resistor. To get the theoretical maximum current you divide the voltage, 20 V in digital operation, with the total resistance , 20 Ohm + 33 Ohm = 53 Ohm.  20 V/53 Ohm = 0.377 A. This gives a total power of 4.7 W in the resistor. I think the real maximum is somewhere between my measured maximum and the theoretical. Thus, a power rating of 3-4 W would be required.

I had a 33 Ohm 4 Watt resistor, but it was too large to fit into the loco. Therefore, I have ordered four 8.2 Ohm 1 W resistors instead. In the mean time, I use three 22 Ohm resistors and avoid to run the loco with extreme load.