The Quarter Wave Stub Resonance and How it Can Kill Your Design
The purpose of this investigation was to see when back-drilling vias in your high speed PCB yields to a significant improvement in the signal integrity of the board. Vias are typically back-drilled in order eliminate stubs and the associated resonance. The back-drilling process is expensive so the purpose of the investigation was to understand when this is economically viable. As a testcase, the USB3.0 (5GBps) protocol was investigated
In order to investigate this, PCB structures encapsulating a long and a short stub were create. These were both analyzed using HFSS and VNA measurements. Both analysis were compared in order to evaluate the quality of the simulations.
Short Stub S-Parameter Measurement Simulation Correlation
RED - Measurement // Green - Simulation
Long Stub S-Parameter Measurement Simulation Correlation
RED - Measurement // Green - Simulation
It was found that little to no benefit would come from the extra cost of back-drilling vias with respect to USB3.0 (5Gbps) . We were able to demonstrate an open eye with a 150 mil long stub.
There are two parameters that you need to be aware of the first is the bandwidth of your digital signal. The second is the quarter-wave stub resonance
As a general rule of thumb, BW[GHz] = 0.35/RT[nsec].
Another general rule of thumb for a stub is that the resonant frequency f[GHz] = 1.5/Len[inches]
In terms of signal integrity stubs do not cause a significant problem as long as the resonant of the frequency of the stub is beyond the bandwidth of the high speed signal.
As Howard Johnson has said, you can design you interconnects so good it might be "so good, it works on barbed wire" https://www.edn.com/so-good-it-works-on-barbed-wire/