Issue link: https://iconnect007.uberflip.com/i/994883
40 DESIGN007 MAGAZINE I JUNE 2018 DDR3/4 Fly-by Topology Termination and Routing Beyond Design by Barry Olney, IN-CIRCUIT DESIGN PTY LTD / AUSTRALIA DDR3/4 fly-by topology is similar to daisy chain or multi-drop topology, but it includes very short stubs to each memory device in the chain to reduce the reflections. The advantage of fly-by topology is that it supports higher-fre- quency operation and improves signal integrity and timing on heavily loaded signals. If you are employing high-frequency DDR4, then the bandwidth of the channel needs to be as high as possible. However, with today's extremely fast edge rates, the sequencing of the stubs and the end termination, and the associate load, can make a measurable difference in signal quality. In this month's column I will look at how best to route DDR3/4 fly-by topology. Reflections occur whenever the impedance of the transmission line changes along its length. This can be caused by unmatched driv- ers/loads, layer transitions, different dielectric materials, stubs, vias, connectors, termina- tions and IC packages. By understanding the causes of these reflections and eliminating the source of the mismatch, a design can be engi- neered with reliable performance. For perfect transfer of energy and to eliminate reflections, the impedance of the source must equal the impedance of the trace, as well as the imped- ance of the load. As signal rise times increase, consideration should be given to the propagation time and reflections of a routed trace. If the propagation time and reflection from source to load are lon- ger than the edge transition time, an electri- cally long trace will exist. If the transmission line is short, reflections still occur but will be overwhelmed by the rising or falling edge and may not pose a problem. But even if the trace is short, termination may still be required if the load is capacitive or highly inductive to pre- vent ringing. Series termination is an excellent strategy for point-to-point routes, one load per net. Figure 1: Fly-by topology for clock, address and command routing.