Been having an interesting time of it setting up some new locos for the layout. The problem appeared to start when I tried to setup the long address for my new Rapido GMD-1. Worked fine at the default (de fault is always de user...) address of '03'. But when I changed it on the programming track to aything else the loco would brick. However, I could do a factory reset to the default address, this worked, and the loco was functional.

Earlier, I had changed the address of a number of units so my first thought was that I sort of knew what I was doing. I had a PTB-100 installed, the only thing that had changed was adding a DCS-100 to the layout. This meant a longer lead to the programming track connections from where my Zephyr was connected. But I did n't think anything of it. Tried the same trick on another loco with a different sound decoder and same problem. Hmmm. Not a feature of the decoder if I can do the same trick and wreck the same damage.

I noticed that when I used DecoderPro to check the settings, every time the address settings changed but were never what I specified. And none of the additional CV changes propagated.

Laid under the thing and using the DT402 throttle, frobbed the address setting a few times and watched the diagnostic LEDs on the board, no error indications. And after a few tries, the four digit address actually made it to the decoder, but none of the other settings did. But while laying there I realized that the gauge of the longer leads might be a factor... reread the docs for the PTB-100 a few more times, then noticed the line that said... r use a dedicated power supply that will provide 500ma at 15 volts DC. Oh... Bet the six feet of 28 gauge wire might have a bit of a problem supplying that...

So... rewired the PTB using heavier gauge wire. The power feed from the command station transformer got upgraded to 16 gauge... the programming track signal leads went to 22 gauge. And short sections of a fatter wire were soldered onto the tiny leads on the PTB. Then, the smoke test...

Needless to say everything now works perfectly. And I am having fun with a programmatic run-in tool, exercising my new loco on a set of rollers.

I find it interesting that none of the news groups or vendor sites said a word about this -- guess it must be too obvious, or subtle. My vote is the later -- everything appeared to work just fine, except it didn't. I am guessing the current flow was sufficient for it to appear to work, but the sound decoders just didn't have quite enough current to actually complete the CV writes -- so I got random values. Interesting.