Differential Strip Line Calculation (USB 2.0)

[Luca82]

Hello,
I'm going to route the D+/- differential pair for USB. I used different simulators (Saturn, Zcalc...) to decide width, gap, PCB stack etc.
My question is: when we create the GND plane on the same layer as the D+/- differential pair, how far should the GND plane be from the D+/- tracks ? Shall it be 10 x W ?
Thanks a lot !

[robertferanec]

- I do not pour GND on signal layers, so I am not sure how far you need to pull it back, but maybe this may help: http://www.icd.com.au/CPW.html

PS: be sure you ask your PCB manufacturer to give you the final track geometry. Do not calculate it by yourself, your calculations will not be accurate. PCB manufacturing process will change some of the values.

[mairomaster]

I've done 0.5 mm clearance around 10 GHz differential pair without problems. That should be more than sufficient for USB2.0. If you have space, better leave too much space than too little.

[Paul van Avesaath]

I agree, more clearance is better, but i definitly use Saturn to calculate the impedance, just to give an idea on what the trace width / spacing would be.. in my case i always let the manufacturer do the final impedance mathcing, but that is because there are a lot of other signals that need it.. it's better to be close than way off.. its that much harder for the manufacturer to fix it..

also in case of only USB i would not be that concerned with the final correct value.. USB is very robust, and getting close to the impedance is enough..

[jankanolv]

Generaly coplanar waweguides will work better with less clearance. Bigger gaps should result in more power loss and more noise in the signal.
In this case (USB 2.0) you could even do without the shielding copper, provided that there is a continious GND plane under the tracks

[Luca82]

Thanks guys for the support ! I'm now routing some Ethernet and mPCIe with microstrip and striplines and I'm using 0.5mm clearance.