Reading time ( words)
Does a 90-degree corner affect signals and should a layout designer be concerned about traces with corners? If they are important, what's a better geometry? Yuriy Shlepnev, Electromagnetics Guru and President of Simberian Electromagnetic Solutions, shows how to analyze the impact of corners on signals and how they affect performance in this presentation.
He demonstrates that bends do affect signals, but so slightly, it is often difficult to measure. To explore their impact, an accurate field solver is essential. Even the small frequency dependence of characteristic impedance in uniform 50 ohm microstrip lines can swamp the impact of the corner. Simbeor, the 3-D full wave field solver from Simberian, allows normalizing out the return and insertion loss of the uniform microstrip to reveal the impact from just the corner.
In this presentation, an 8 mil wide 50 ohm microstrip is analyzed, with and without a 90-degree bend. Even at 25 GHz, the reflected signal is less than -25 dB, or less than 5% reflected back. This is a small, but easily simulated discontinuity.
Though this is already pretty small, if you wanted to reduced it further, a common technique is to chamfer the corner with two 45-degree bends. But what's the optimum length of the chamfered section--does it even matter?
Simulating the return loss for different chamfer dimensions shows an optimum value of about 1.2 for the ratio of the chamfer length to the line width. This drops the return loss at 25 GHz from -25 dB to -46 dB.