Don't Breathe Too Hard
By Dr. Eric Bogatin
www.BeTheSignal.com
August 2007
High-speed serial links are fragile. You have to worry about tiny vias that might make up less than 1 percent of the total path length. You have to worry about the size, shape, and orientation of the glass weave in the laminate. You have to worry about the surface roughness of the copper. You have to worry about the size and location of the via clearance holes in the copper planes near the connectors. All of these delicate adjustments must be just right for a successful design. Now there’s a new worry.
In a paper originally presented at the IPC Expo 2007 Technical Conference, and recently printed in the June 2007 issue of Printed Circuit Design and Manufacture Magazine, Gary Brist, Guy Barnes, Jr., and Jason Schrader from Intel and Paul Hamilton from Tyco describe another sensitivity for high-speed serial links: humidity.
In their paper, “Humidity-Dependent Loss in PCB Substrates, the authors describe a series of measurements performed on circuit boards manufactured from a variety of resin-glass combinations. One set of boards was manufactured with Rogers 4350 material and others were manufactured with FR4 resin and various glass weaves such as 1080, 7628, and 2116.
Test boards were constructed as eight-layer boards using these and other sets of materials. Both microstrip and stripline geometries were used. Though microvias connected the buried signal lines to the top surface where they were probed, precision clearance holes were patterned in the return planes of the stripline structures to emulate arrays of through-hole vias. Figure 1 shows an example of the top view of one microstrip test pattern.
Figure 1. Top View of Microstrip Test Pattern
The authors report that the distance between the signal traces and the top layer and the number of open holes in the planes strongly affected the time constant for water diffusion and a resulting change in the dielectric properties. Surface traces experienced equilibration times of hours, while equilibration for buried traces took days to months.
Brist and his team measured the impact on dissipation factor and dielectric constant from humidity and temperature over a period of more than five months. From the time dependence of the change in dissipation factor, they were able to estimate the diffusivity of water into the material.
They note that the diffusivity depends on the temperature and humidity as well as the type of glass and the open area in the planes. Even so, their value of the diffusivity of water into FR4 at 38 degrees C and 95 percent RH as 1.31 x 10-6 mm2/sec compares well to the diffusivity at 85 degrees C and 85 percent RH of
1.51 x 10- 6 mm2/sec reported elsewhere, based on weight pick-up measurements.
The dissipation factor of the laminate was indicated by the measured insertion loss, or attenuation, of the transmission lines. Using a switching matrix to connect multiple samples remotely to the VNA enabled samples to be measured in-situ in an environmental chamber at elevated temperatures.
Figure 2 shows the measured attenuation per length at 5 GHz, across a 15 to 85 degree temperature range. The bottom curve represents when the laminate sample was baked and measured dry. The top curve represents the laminate after the sample was conditioned at high humidity. Higher loss (more negative value) indicates a higher dissipation factor.
Figure 2. Measured Attenuation Per Length at 5 GHz
The authors state that the as-received sample showed a loss-per-length of approximately -0.95 dB/inch at 25 degrees C. After baking to remove all moisture, the loss, at the same temperature and frequency, was reduced to -0.65 dB/inch. When saturated with moisture, this same sample showed a loss of -1.55 dB/inch.
This is a loss of almost 2.5x from dry to saturated. Furthermore, when saturated, the loss was temperature dependent, varying from -1.55 dB/inch to -2.4 dB/inch, a factor of 1.5x increase.
As bit rates enter the multi gigabit-per-second range and proliferate into consumer products, such as with SATA, PCIe-I, and PCIe-II buses, the attenuation from dielectric loss can dominate performance. If a system is designed for operation assuming one value of dissipation factor, for example, based on measurements in Phoenix in the wintertime, and the product is used in Kansas in the summertime, the attenuation of signals from a dissipation factor increased by moisture and temperature may collapse the eye and cause an unacceptable bit error rate.
With a diffusion equilibration time that varies from trace to trace and board to board, debugging such problems may be tricky unless designers are aware of this problem and consider its impact. The authors conclude that since temperature and humidity affects transmission line loss, the expected environmental conditions must be considered during system design and validation.
This and other signal integrity topics are covered in Eric’s public classes and online lectures, available from his website, www.BeTheSignal.com. Send your signal integrity technical questions to DoctorIsIn@BeTheSignal.com.
Bio: Eric is president of Bogatin Enterprises, whose mission is to set the standard for signal integrity training. He is the author of Signal Integrity - Simplified, published by Prentice Hall. Check out his public signal integrity classes posted on www.BeTheSignal.com. He can be reached at eric@BeTheSignal.com.
|
Literature
> 
