Connections between grounds

GND loops are nasty things. connecting GNDs where voltages and levels have to be independent may cause big problems. Not connecting GNDs can also have desastrous effects depending on situation.

One point of connection is safe. If there is no othe rcommon point, there is no voltage 8as voltage is always a difference and a diference requries a second connection as reference), but with only one connection, no current can flow and no signals can be submitted.

For critical applications, there are USB cables with optocoupling. The expensive ones have also an isolated DC/DC transformer so besides the signal delay (and the fact that some dirty tricks like USB charger identification through A/B signal lineimpedance won't work) these are a really good thing if you fear potential problems. However, they are rather expensive.

The main question is: what are your different voltages sources and where might be hidden potential differences. Poor PC power supplies may make an AC or even DC conneciton between USB GND and AC power side. THis is harmless until you have a second device that also has a connection to AC side, but maybe with a difference in earth level or phase. Then cross-currents may flow and can destroy your electronics.
I've also seen devices with a cheap capacitor supply, or where the DC electronics was 'floating' on the AC line, depending on how you plugged it in (this was a processor controlled 20-flash stroboscope where the flash capacitors were charged from 220V AC and the electronic did the trigger - without any isolation. It worked until I tried to add an external switch)

Another thing are inductive currents (you said there is a motor somewhere?).

Without knowing the schematics, it's difficult to pinpoint a reason.

Mikolaj Filar said:

1. Voltage on transformer: 17 V DC, 0V AC  -> expected

This is typical in unloaded non regulated  power supply. It is not a problem for the driver chip. For the stepper you may want to add series resistor with the motor windings to limit the voltage.

Mikolaj Filar said:

2. Voltage between transformer GND and Launchpad GND: 0 V DC, 20 V AC

This AC voltage is just minor capacitive coupling. If you add 1 kohm load at  your meter, you will see that this signal has less then 1 mA of current. (<1VAC)

Your driver chip ground (0 VDC) should be directly connected to your launchpad (0VDC) ground.

Peter

Peter Dvorak said:

This AC voltage is just minor capacitive coupling. If you add 1 kohm load at  your meter, you will see that this signal has less then 1 mA of current. (<1VAC)

Indeed, this is a typical thing with switching supplies. To keep the overtone oscillaitons during the swithcing under control, tehre ususally is a 1nF cap between primary and secondary side. We have an expensive (listed at ~ $40)  switching wall-wart supply (6V/2A) from HP which has ~ 90VAC between output and earth potential. so the maximum current is 90nA AC (assuming 1nF cap, but maybe more), bu tit still can be enough to fry an input pin.

PC USB GND is often connected to earth, sometimes directly (through the shield and case of the PC) or indirectly (through a similar cap in the PC supply which worst case won't be in phase to the other).

If you have both, signal and GND, connected between the two sources, this is normally not a problem, as the same AC overlay appears on both, GND and signal. It should be considered anyway as a possible source of problems.