USB Data Only for data acquisition ('scope)

[theoldwizard1]

In order to minimize "noise", is limiting power (likely not used anyway) and ground between the "host" computer and the data acquisition device a good or bad idea and why ?

To keep it simple, let's limit the discussion to USB 2.0.

[singapol]

Actually it's a wide ranging subject touching on circuit design and layout and hardware (power supply).
It is well known anything PC related is noisy despite the hype in terms of data acqusition. It is analog signal in real life. Depending on the type of signal small or large aquired using a linear supply saves you headaches.
Then take the chance as it gets to PC unmolested. In theory having host PC and data acquisition device on same ground is prefered as the ground noise is referenced to same ground. For a more detail treatment
try to download  an old copy of Henry Ott's Noise Reduction Technique in Electronic Systems 2nd. edition.
There are tons of articles on noise online.

[theoldwizard1]

So, imagine a couple of volts applied to D+/D- on your USB.

Excuse me but how does a "couple of volts" wind up on a differential serial bus ?  I thought the whol point of a differential transceiver was to isolate it from power and ground ?

[Ian.M]

USB isn't fully differential. It uses a single ended zero state (SE0) with both lines low for various signalling purposes.  As such it cant tolerate any significant difference in 0V potential between the device and the host.

[David Hess]

Noise is usually less of a problem than destructive overload.  Accidentally connect the probe ground of your USB instrument to AC hot (or AC neutral if the neutral connection fails) and you may end up replacing the PC and various devices attached to the PC.