While changing the connections to allow switching "Remote" pin (17) ON and OFF, I must have changed a bit the breadboard orientation or position and began getting CRC errors again.
Now I am sure that my measurements of "3V low" are wrong, and that the voltage is more like 1V. So taking this hypothesis, I thought that in this case, the divider could be much more balanced (like 10K against the internal 5K)(*) and the tests showed that it improved things very much : I ran a full "intense skipping" session and got zero CRC errors (see attached file -
Edit : an Excel formated version is available in
this post).
And this wrong level explains many things :
- the "strange level"
- the observations of Mox (thanks again for having had this walk with your scope :-))
- the fact that CristianC could connect the pin directly to a PC. Indeed, at 1V/12V, the use of the MAX 232 is questionable for most recent PCs that have their threshold above 1V. I would like to hear the opinion of people used to serial communications (Ale again ?) about the removal of the MAX3232 altogether... Only for data coming from the PC probably would it be wiser to change levels I think (at least to avoid sending negative voltage to the CDC)
(*) in fact, the divider could probably be removed too. If we go on that way, the whole interface will soon look like a few wires :-)