Study of the insertion loss of a differential pair of through holes for the 25-Gbps serial interconnect

To design future high-speed interconnects that operate above 25 Gbps, accurate scattering parameters (s-parameters) of structures, such as through holes, are necessary. We estimated the s-parameters of a differential pair of through holes by using the de-embedding technique to eliminate undesirable loss and delay in the feed lines connected to the through holes. By comparing this insertion loss and the estimated insertion loss of a corresponding electromagnetic (EM) simulation model, we found that the distance between the center of the drill holes and the center of their clearance could be a reason for the increase in the insertion loss of the through holes in addition to the open stub effect above 20 GHz. We also discuss the reason for the error in the de-embedding from the viewpoint of measurement repeatability. We estimated contact resistance, parasitic capacitance, and parasitic inductance at the air coplanar (ACP) probe tips from the measured s-parameters. The result shows that these parameters varied about 0.1 Ω, 0.1 nH, and 0.05 pF due to reasons such as the contact condition or the flatness of the PCB. To mitigate these parasitic lumped elements, probes with smaller tips are preferable when we measure the s-parameters of structures for de-embedding.