Design007 Magazine


Issue link:

Contents of this Issue


Page 45 of 65

46 The PCB Design Magazine • November 2016 current is forced into a small width of copper. This increases trace resistance, inductance and skin-effect losses. Contrary to common belief, closely coupled pairs do not improve EMI. This is because it is the common-mode signals from the drivers—the natural imbalance (skew)— that radiates. In conclusion, to minimize radiation and crosstalk, of a differential pair, one must think explicitly about the common-mode component of the signal. It is not a matter of which is bet- ter—tight or loose coupling—but rather which scenario better avoids timing skew that creates this common-mode signal. It is imperative to determine exactly where the differential imped- ance levels-off. But without a good field solver that simulates signal coupling and flight time, you are really just taking a stab in the dark, which is not good design practice. Points to Remember • Conventional wisdom tells us that tight coupling between the signals is better than loose coupling, as it reduces undesirable cou- pling/crosstalk from aggressor signals. • On a PCB, the return current path of each trace of the pair flows directly below each trace in the reference plane. • If a differential pair is not perfectly bal- anced, then the degree of cancellation is not determined by the spacing, but rather by the common-mode balance. • Most digital drivers have poor common- mode balance and therefore differential pairs often radiate far more power in the common- mode than in the differential-mode. • For a perfectly balanced differential signal, the radiation from one trace exactly cancels the radiation from the other as they are equal and opposite. • The common-mode signal represents an average of the two signals in a pair. The radia- tion is identical on both traces and therefore it does not cancel but rather reinforces. • To minimize radiation and crosstalk, you must think explicitly about the common-mode component of the differential signal—skew cre- ates this common-mode signal. • To minimize the skew, match the electrical lengths and correct any shift immediately, after it arises, by adding length (hence delay) to the shorter trace. • Time-delay skew can also be introduced by a variation in the dielectric constant of the glass-resin composite. • The transformation from differential to common-mode also takes place on bends and non-symmetrical routing near via and pin ob- structions. • Mirror symmetry, about an axis along the interconnects, avoids mode transformation. The symmetry property preserves the signal in the differential-mode which does not radiate. • The differential impedance of a pair, levels- off at a particular coupling point. Beyond this point, the differential impedance will always be the same regardless of increased spacing. • As the traces get closer, both differential and common-mode impedances are reduced so the traces must be made narrower to compen- sate. • Narrow traces increase trace resistance, in- ductance and skin-effect losses. • Contrary to common belief, closely cou- pled pairs do not improve EMI. PCBDESIGN References 1. Barry Olney, Beyond Design, "Differential Pair Routing." 2. Eric Bogatin, Fundamental Myth-Concep- tions. 3. Lee Ritchey, Right the First Time. 4. Yuriy Shlepnev, Analysis of differential line transition from tight to loose coupling, and practical notes on mixed-mode transformations in differential interconnects. 5. Howard Johnson, High Speed Digital De- sign. Barry Olney is Managing Director of In-Circuit Design Pty Ltd (iCD), Australia. The company is a PCB design service bureau that specializes in board-level simulation. iCD developed the iCD Stackup Planner and iCD PDN Planner software. Visit UNCOMMON SENSE

Articles in this issue

Links on this page

Archives of this issue

view archives of Design007 Magazine - PCBD-Nov2016