Electronic – Stability of USB +5V output

It sounds like your device needs more than the 100 mA it is guaranteed, but does not enumerated properly to request the extra current. You can get up to 500 mA from a USB port if you ask for it and the host grants it.

There is no guarantee that the host will grant the request, but desktop machines pretty much always do since they have plenty of power available. In fact, the desktop motherboard schematics I have seen show only a polyfuse between the 5V supply and each USB port. There is no active electronics limiting current. On those machines, you get the full 500 mA even if you don't ask for it.

Laptops however have limited power and therefore actively monitor and control USB power in most cases. Some machines may limit the sum of all USB devices to a certain current, or limit it more when running on battery only, or when the battery is low, etc. Note that USB powered hubs can't ever grant the full 500 mA since that is all they are getting themselves.

The reason your device won't enumerate anymore is because you broke the ground connection. All three lines D+, D-, and ground are required for communication. You can cut the power line if the device is self-powered. Self powered devices generally have no connection to the USB power line. That's perfectly OK, but they still have to connect via ground.

1. You are correct about the floating nature of the wall wart output.
2. There's no harm in leaving USB +5V unconnected. On the other hand, you might have to connect it to your controller, because that's part of its scheme for detecting the USB connection.
3. Connecting USB GND is necessary. D- & D+ carry a differential signal, however the common mode needs to be within the operational range.
4. Self-powered USB by itself doesn't required galvanic isolation. Whether or not your instrument requires galvanic isolation (and where to have it) depends on the application (?) of your instrument.

A sketch of a self-powered USB device can be found here see 6.2 on p.24.