Firstly the 2mm scale 04, mentioned before in the blog. This is a Farish model, which has 2mm scale wheels fitted to it (modified 2mm Scale Association castings and I have a lathe!). It finally has the coupler fitted to the front and it works nicely.
I did encounter a couple of problems along the way, one magnetic and one software.
The magnetic problem is quite simple - the Farish motor is in the bonnet, so near the front of the loco there is some magnetic field from the motor. This proved enough to just cause the magnet in the coupler to "stick" in the open position. The eventual solution was a larger counterweight on the shaft. Video below shows operation, and the counterweight. The filling of the big gap in the bufferbeam is not yet complete.
The software problems were a bit more fraught, its yet another undocumented feature/bug in CT decoders. Without getting too far into CV numbers, it revolves around CV152 and CV153 in decoders with version 56 (DCX74) or 59 (DCX75) firmware (might be other numbers below 59 as well). In theory those CV's specify the output wire to activate during an uncoupling sequence. A value of "2" works correctly for the white wire. The documentation (German language versions) say its a "bitwise" allocation, so in theory "4" would work for the yellow wire.
Yellow does not respond to a value of "4". However, a value of "1" does work for Yellow, but with a minor glitch.... If the loco is set for Yellow at one end, and White at the other, then the sequence to operate White will also cause Yellow to momentarily activate whiles the loco changes direction. Its not a massive issue, but it might cause some systems to disconnect unintentionally whilst operating the uncoupler at the other end of the loco.
In the end, I rewired the loco so that both coils are in series, rather than independent function outputs. With the rotating action, the "wrong" end will uncouple and re-couple, so there is no disadvantage from operating both together, and arguably a little bit of wiring simplification.
Further bench testing has shown that a version 66 decoder (DCX74) has yet another change to CV152/153, which looks like a bug fix. The values are now "White=1" and "Yellow=2" (logical!) and the behaviour is correct at both ends. For testing, I use a home made decoder tester, which is a collection of LEDs and resistors connected to a small screw terminal block. The decoder motor and function leads can be connected to these, the decoder red/black go via crocodial clips to the command station/programmer. In my case a Sprog which is driven from my computer using JMRI. Using JMRI its quick to flip between different CV settings and observe the results.
So, to the recommendation for anyone else trying this stuff; if using a CT decoder, check how CV152/153 works in advance for your firmware. If in doubt, use the White output wire.
I'd like to see reports from others using the CT decoder and CV152/153; what happens with your decoder, what is its version number. It will help with development of the JMRI decoder definition files.