How the Fiber Weave Effect Influences High Frequency Signal Integrity
The fiber weave effect can wreak havoc in networking equipment for copper and fiber
When you look at any material under the microscope, it can feel like looking into another world. I spent many long hours on SEMs and TEMs comparing different materials and examining changes in optical and optoelectronic devices during operation. With new devices running at ever higher data rates, designs on rigid substrates need to accommodate one inconvenient phenomenon: the fiber weave effect.
This doesn’t just refer to one effect; there are multiple signal integrity problems that can arise due to the fiber weave effect. At sufficiently low edge rates (> 1 ns) and frequencies (< 1 GHz), you would probably never notice effects from fiber weaves styles. The fiber weave effect rears its ugly head once signal frequencies and bandwidths become high enough to accommodate applications like 100G/400G, mmWave devices, and ultra-high speed SerDes designs. Fortunately, there are some steps you can take to account for the fiber weave effect during layout and routing.
How to Think About the Fiber Weave Effect
All resin/glass based PCB laminate materials are produced with a loom, which is used to create a glass weave as a reinforcement in a PCB substrate. Newer materials that are specialized for high speed/high frequency designs, such as a recently released laminate from Rogers Corp., are being optimized to have low losses and desirable CTE, Tg, and thermal conductivity values. However, little can be done about the weave style and the signal integrity problems they create. Even the most advanced resin-based laminates are optically inhomogeneous, anisotropic materials, meaning their dielectric properties vary in space and along different directions.
As a designer, you cannot necessarily plan ahead to eliminate fiber weave effects. You can certainly specify a desired orientation for a trace with respect to the fiber weave arrangement, but the sizes of traces compared to cavities in a PCB substrate make it difficult to predict exactly where your traces will run along the board. This means you need to think in terms of average dielectric constant, and thus in terms of average propagation delay.
Fiber weave styles. Loose weaves (left) create greater skew and impedance variations in a board compared to a tight weave (right). Image credit: Chen et al. (MDPI).
How Fiber Weave Style Affects Skew
Because cavities in the fiber weave are formed by gaps between glass bundles, traces routed over these cavities will see a different dielectric constant compared to the glass bundles. The difference in dielectric constants can reach a factor 2, depending on the materials used in the substrate. Using the dielectric constant to calculate skew may not always be accurate as the dielectric constant value quoted on datasheets depends on the measurement technique. In any system where control over timing skew is critical, you need to ensure the value you are using for your propagation delay calculation is accurate. This is quite important when examining differential pair skew as the spacing between pairs can cause each pair to see different single-ended skew values.
One formula that can be used to determine a skew involves using the dielectric constants of the glass and resin weave. Skew basically accumulates due to the difference in propagation delays across each material, which is proportional to the difference in dielectric constants:
Skew approximation equation
As always, test out datasheet values and any calculation for yourself using a test coupon! Also take a look at this recent publication for some experimental data gathered with different fiber weave styles.
As was shown in a recent publication in Signal Integrity Journal, routing at a slight angle with respect to the weave pattern can reduce timing skew (standard deviation) from ~7 ps/in. to less than 1 ps/in. Note that this is solely for skew due to the fiber weave effect; other sources of skew like clock jitter and delay mismatch still need to be considered. However, angles involved were only ~0.04 rad, equivalent to ~2.3 degrees. In other words, the skew standard deviation can be reduced by approximately 3 ps/degree, up to a maximum reduction of ~7 ps.
Skew reduction Image credit: Bogatin et al. (Signal Integrity Journal).
Fiber Weave Resonances at GHz Frequencies
The cavities in loose fiber weaves can also act as partially open resonators, and resonances generated in fiber weave substrates are typically ignored. One must remember that the electromagnetic field is not confined within a trace; it actually exists around the trace and is confined in surrounding media. This means a travelling high frequency signal, or a digital signal with large bandwidth, can easily excite one or more standing wave resonances in these cavities.
The lowest order fiber weave resonant frequency is typically ~50 GHz for loose weaves (e.g., 45 GHz for a 60 mil weave pitch in FR4), putting them decidedly in the mmWave regime. These resonances can then excite sub-harmonic cavity resonances through resonant coupling. In other words, the fiber weave pockets, nearby conductive structures, and the parasitics created by each act as a source of radiated EMI. This particular issue was recently discussed in Signal Integrity Journal.
Strong resonance in these cavities can also couple inductively or capacitively into nearby circuits. This coupling is more of a problem in RF signal chains involving power amplifiers, high power FET drivers, and similar circuits that produce strong RF fields. This effect appears as a drop in the insertion loss profile at successive fiber weave resonances. You can measure this effect by extracting the S-parameters from a test coupon with a vector network analyzer.
In summary, if you want to prevent problems with resonances and insertion loss dips, aim for the tightest glass weave style that meets your loss, CTE, Tg, and thermal conductivity requirements. A tighter weave style will generally have higher frequency resonances, although there will be definite tradeoffs that need to be balanced. Accurately accounting for skew and ensuring controlled impedance requires determining the right average dielectric constant to use in your impedance calculations. In the event cavity emissions become problematic, you might consider using a conformal coating as a shielding material.
The layer stack manager in Altium Designer® allows you to define the average dielectric constant your signals will see as they travel along a signal trace. This makes it an ideal tool for compensating skew from the fiber weave effect in your board. The post-layout simulation tools are also useful for examining crosstalk between traces carrying high frequency signals and for controlled impedance routing. You’ll have access to an extensive library of standardized materials and weave styles that you can use in your stackup.