IEC: On-Line Education: WPF: Design Considerations for Gigabit Backplane Systems

System attributes evaluated in this study include line length, trace width, trace topologies (single-ended, broadside differential, and edge-coupled differential), material losses (for FR–4 and more exotic, lower-loss materials), conductor losses, plated-through hole (PTH) launches, routing considerations, and board-to-board interconnects. The effects of the aforementioned components or subsystems for backplane systems are quantified to present possibilities for a viable system based on fast data rates. 2.5 Gbps was chosen as the data rate for sample system comparison. These trends would also be useful in understanding design parameters for other data rates.

Analysis of individual components will verify the 2.5-Gbps data stream in order to create a complete system. Once the test system is analyzed, correlation between the system measurements and a mathematical combination of individual components will use all data for additional system descriptions. Time and frequency domain analyses are in the form of reflection data (time domain reflection [TDR], S11) and transmitted data (time domain transmission [TDT], S21, eye pattern). Test equipment includes a 3–Gbps HP (70841B) pulse pattern generator, a Tektronix 11801B TDR, and an HP 8722D 40 GHz vector network analyzer (VNA).

As signal speeds increase, system performance is limited by long lengths, impedance mismatches, and various noise in the system. Long lengths relate to material losses, conductive losses, and greater distances for noise interjection. Because the system is more sensitive to these factors, engineers implement different techniques designed to combat the aforementioned nuisances. Different material laminates may alleviate dielectric losses and dispersive effects. From a table of different high-performance laminates, nine laminates were chosen for board studies and highlighted in Table 1.

Material Type Availability Er Df Tg
(degrees C)

Nelco N4000–13 all types 3.95 @ 10 Ghz 0.01 @ 1 Ghz 210

N6000 all types 3.3 @ 1 Ghz 0.005 @ 1 Ghz 210

Arlon 25N 0.006" increm. 3.25 @ 10 Ghz 0.0024 @ 10 GHz > 225

Megtron R57 15 all types 3.5–4.2 @ 1 Mhz 0.010-0.015 @ 1 MHz 180

Allied Signal FR 408 all types < 3.6 50 Mhz–1 Ghz < 0.009 50 Mhz–1 Ghz 180

Gore Speedboard C prepreg only 2.2.–2.6 @ 1 Mhz 0.003 @ 1 Mhz 190

Rogers 4000 series 0.0033" increm. core 3.38 @ 10 Ghz 0.004 @ 10 Ghz > 280

0.0040" increm. fill

Polyclad PCL–LD–621 all types 3.5 @ 1 Ghz 0.006 @ 10 Ghz 190

GIL GML 1000 0.020", 0.030", 0.060" 3.05 @ 10 Ghz 0.003 @ 10 Ghz 135

MC 5 cores/prepregs 3.26 1–15 Ghz 0.0015 @ 10 Ghz 145

G.E. GETEK all types 3.6–4.2 @ 1 Mhz 0.010–.015 @ 1 Mhz 175–185

ISOLA GIGAVER all types 3.5–4.0 @ 1 Mhz 0.003 @ 1 Mhz 210

Taconic RF 35 0.0035" increm. 3.5 @ 2 Ghz 0.0018 @ 2 Ghz 315

cores only

Table 1. High-Performance Printed Wire Board (PWB) Materials

To combat noise interjection, device manufacturers are implementing differential pairs.

2. Broadside versus Edge-Coupled Differential Pairs >>