Is angle routing recommended for XAUI??

Do you mean 45% champher on bends?

Nobody can answer that question, there are too many other variables, like how long is your trace, how good is your impedance control, what are your geometeries within each pair? Even how many bends you're introducing.

IF your impedance control is excellent, your traces very short, and you have little crosstalk in your nets, then you will have room for a bit of noise introduced by 45deg or 90 degree bends.


Angled corners radiate more than curved ones (which means impedance discontinuities and SI issues too). There is also more skew introduced.

Whether you can live with this or not is up to you. Lay it out, run a sim and see.
If you're not sure, curvy traces are better for SI. Whilst it may tke you longer, you know that this particular aspect of layout can't be improved.

I guess what I'm saying is do the SI work. If you don't have the resources, then you need to balance the cost to your schedule (to add rounded corners) with the risk of getting a card back with "issues". I know nothing about XAUI, will it autonegotiate to a lower speed? Is the card a development prototype, or will this ship to a customer?

I don't think there are hard and fast rules. You need to decide based on your application.

Do you mean 45% champher on bends?

Nobody can answer that question, there are too many other variables, like how long is your trace, how good is your impedance control, what are your geometeries within each pair? Even how many bends you're introducing.

IF your impedance control is excellent, your traces very short, and you have little crosstalk in your nets, then you will have room for a bit of noise introduced by 45deg or 90 degree bends.


Angled corners radiate more than curved ones (which means impedance discontinuities and SI issues too). There is also more skew introduced.

Whether you can live with this or not is up to you. Lay it out, run a sim and see.
If you're not sure, curvy traces are better for SI. Whilst it may tke you longer, you know that this particular aspect of layout can't be improved.

I guess what I'm saying is do the SI work. If you don't have the resources, then you need to balance the cost to your schedule (to add rounded corners) with the risk of getting a card back with "issues". I know nothing about XAUI, will it autonegotiate to a lower speed? Is the card a development prototype, or will this ship to a customer?

I don't think there are hard and fast rules. You need to decide based on your application.