When someone got the same problem with the RAMPS-FD board (meaning, the Due which has the ground reference has the issue), they suggested 1kohm.
I think you don't need to go that high. Even as low as 100 ohms may be enough to stop the problem.
Really though, the issue is the ground on the Melzi (and any board that has its own USB connector) is that the USB shield tends to be wired directly to the DC ground on the board. Using a resistor (even a small one like 100 Ohms) between the shield of the USB and the ground might solve this problem. That would of course require a redesign of the board to do it properly. Also while this "should" be fine, it needs to be tested, as we need to be sure the issue is with the USB shield, or the USB ground pin (which may be different again).
FYI: The main problem here is that most Arduino's, their clones, and the reference design, all expect that you'll be running them in the following scenarios:
1. Drawing all your power from the USB
2. Drawing all your power from a DC plugpack, which has no AC earth grounding.
3. Running the device completely stand-alone when in operation, and completely disconnected when programming.
As such, they haven't bothered to really worry about USB ground loop issues caused by wiring DC ground to AC ground at both ends. And therein lies the issue.
I think you don't need to go that high. Even as low as 100 ohms may be enough to stop the problem.
Really though, the issue is the ground on the Melzi (and any board that has its own USB connector) is that the USB shield tends to be wired directly to the DC ground on the board. Using a resistor (even a small one like 100 Ohms) between the shield of the USB and the ground might solve this problem. That would of course require a redesign of the board to do it properly. Also while this "should" be fine, it needs to be tested, as we need to be sure the issue is with the USB shield, or the USB ground pin (which may be different again).
FYI: The main problem here is that most Arduino's, their clones, and the reference design, all expect that you'll be running them in the following scenarios:
1. Drawing all your power from the USB
2. Drawing all your power from a DC plugpack, which has no AC earth grounding.
3. Running the device completely stand-alone when in operation, and completely disconnected when programming.
As such, they haven't bothered to really worry about USB ground loop issues caused by wiring DC ground to AC ground at both ends. And therein lies the issue.
