[time-nuts] Bodnar "Precision Frequency Reference (GPS Clock)" AND LeoNTP Networked Time NTP Server Questions

jimlux jimlux at earthlink.net
Fri May 18 09:59:03 EDT 2018


On 5/18/18 6:09 AM, Clint Jay wrote:
> Interested to know how much noise would be from USB signalling and how much
> is " machine noise"  from the PC as my understanding of USB signalling is
> that it's differential so such should be low noise?

Every device I've worked with that had a USB interface has radiated 
significant power around 12 or 24 MHz. (I've not worked much with 
devices that support 480Mbps USB3..)

Indeed, the data wires are twisted pair (although the "twist rate" in 
most cables is pretty sketchy), but the signaling is voltage mode, not 
switched balanced current like LVDS.  So I think you get significant 
current spikes on the power wires as you charge and discharge the 
capacitance/inductance of the (unterminated) data lines.


The shielding is, in a lot of consumer devices, common to the power 
supply negative lead.  So the shield carries part of the power supply 
current, and then radiates.

In theory, USB devices sold in the United States should meet FCC Part 15 
(and the similar requirements in the EU and elsewhere) - but I'll bet 
there's a LOT of stuff out there that isn't tested, nor would it pass.


High Speed USB might be better for EMI/EMC - you're just not going to 
push 480Mbps through an unterminated system with crummy noise properties.

TI has an application note
https://www.ti.com/sc/docs/apps/msp/intrface/usb/emitest.pdf



More information about the time-nuts mailing list