I'm fairly new to Arduino programming but have a good deal of experience with low-level programming early processors, having originally cut my teeth on Z80, 8088, etc. I'm looking at developing a sizeable project using a Mega 2560. The project includes quite a lot of off-board logic, leds, and an RPi. Thus I don't want to overload the 2560 voltage regulator and really want to use common rail 5v power with a switch-mode supply. However, I also need debugging / console access, so this gives me the immediate problem of direct 5v driving whilst the board is being driven with 5v from the USB port, which is a no - no. The solution appeared to be fairly simple - use a CH340 - based USB to serial converter and connect it to RX0, TX0 and Reset (DTR output connected via a blocking capacitor) but with its 5v output disconnected. With this setup, the Arduino communicates quite happily with the IDE-based serial monitor (the sketch was downloaded via the on-board usb port), but on trying to download a new sketch via this route, AVRDude fails with a timeout. At the commencement of the download the processor is being reset, but the bootloader doesn't appear to hansdshake with AVRDude (properly). Is this a known feature or have I missed something?
Problem solved. The standalone CH340-based converter for whatever reason isn't man-enough to reliably pull down the RX0 line when connected to the Mega 2560 (also with a CH340 converter on board). I've seen suggestions posted in various places that this can be a problem with respect to DTR (though I don't have a problem with it). My solution is to insert a 7417 open-collector non-inverting buffer between the output of the stand-alone converter and the RX0 input on the 2560 board. I've not been able to find a schematic for the Mega 2560 board to determine its pull-up resistor size but imagine it must be relatively small. Thanks to those who made suggestions.