The 'tests' linked above, show nothing useful whatsoever.
For a start, the USB ports are unloaded. Loading a rail with a fixed current draw is the only way to view ripple and draw meaningful conclusions. The various USB 'ports' are likely under complete control of the OS and being cycled/polled on/off at high speed or in power saving mode as there is no attached device pulling current.
The scope is not isolated from the ground of the PC/Laptop and the computers may or may not be effectively grounded. In short, it's just a whole lot of meaningless screen grabs.
I'm happy to test a bunch of USB ports from a pile of PCs/Laptops on my DSO and post the scope shots, but let's determine a current draw applicable to actually load the rail effectively.
Did some loaded USB measurements too, last weekend, will forward'em soon to Archimago to post updated results. On the 8-bit FFT (bandwidth <100kHz) the only difference was around 1kHz...a small spike, but the rest of the graph was basically the same. By loaded I mean: USB port connected to PLAY or to EssenceOne.
Oh...scope was completely isolated, I've used a MBP on it's battery, thank you. Also, I've used a spire on the GND's probe, just to increase the accuracy of the measurements.
However, if someone thinks the process can be improved, please let me know how.
Thank you,
Raul.