Micro-machining of trenches to form shielded transmission lines: a laser and metalization method for forming a shield around a conductor, permitting 10 Gb/s signals with no crosstalk or EMC.

The introduction of laser micro-machining in PCB manufacturing has led to cavities or non-round "features." These features can be used for via holes and embedded components. Another idea is the formation of electromagnetic shielding. By cutting a groove/trench parallel to the transmission line and filling it with a conductive material connected in ground, a shielding can be formed.

With an increase in frequency, signal sensitivity and EM radiation increases, a problem for digital and analog/RF systems. Shielding transmission lines can solve this. Similar to a coaxial cable, an embedded shielded transmission line in a circuit board can suppress the noise into and from the transmission line, making it a perfect environment for sensitive signals; e.g., analog/RF and high-speed (10Gb/s) digital.

The constraint in building shielding is the processing of the trenches, and recalls discussion from the mid 1990s on which technology was best for forming microvias. The processing time required for cutting trenches depends on the length and depth of the shielding, similar to microvia drilling. Alternative technologies such as plasma etching and photodefinable dielectrics were never adopted on a large scale by the high-end market due to technical limitations. One could argue the use of plasma etching or photo formation would form the trenches in a fraction of the time of a laser, although the problem is a bit more complex.

The construction of a shielded transmission line requires three layers (FIGURE 1). The signal is positioned between two ground planes; the two grooves on either side of the signal are closing the cage forming a 360[degrees] shield similar to a coaxial cable.

[FIGURE 1 OMITTED]

In comparison to a typical microvia depth (50-100 [micro]m), the depth of the trench is typically 300-500 [micro]m. The depth is determined by the objective to form 50 [OMEGA] impedance-controlled lines, which implies a certain thickness. Skin effect losses, which become significant in the GHz range, drive the use of traces of a minimum 150-200 [micro]m.

Cutting "clean" trenches in ~400 [micro]m-thick glass of ceramic reinforced material using a laser is difficult. This is the biggest challenge for bringing this technology to market. As it is already difficult to cut a "clean" microvia without glass protrusion, cutting deeper trenches in multiple glass/ceramic layers is a challenge.

As such, materials such as resin-coated foils have been phased out. RCF (resin-coated foil) has low thickness (~50-100 [micro]m), is sensitive to cracking and is unproven for press-fit technology, (as is required for shielded transmission lines), and for sequential processing. A better alternative is DuPont's Thermount. This material can be built in a number of thicknesses, has a reinforcement and is available in B- and C-stage versions. The first shielded transmission line products were built with Thermount due to these advantages and ease of laser processing. This material does not contain dense areas as in glass weaves, i.e., glass knuckles. A number of shielded transmission line products have been built using Thermount and function well.

High-speed signals (10 Gb/s) require lower-loss materials (<0.001), unless technologies for digital-signal boosting and enhancement such as pre-emphasis and active equalization become mainstream. Thermount, FR-4 and Rogers 4000 have been selected as materials for forming trenches.

Depending on the type of laser used, different results can be obtained. The wavelength is the most important factor, followed by the pulse-shape, duration and frequency. FIGURE 2 shows a section of a transmission line formed in Rogers material. Glass protrusion is visible by the copper knuckles on the trench wall. The plating process must be capable of covering the fibers to avoid voids in the shielding.

[FIGURE 2 OMITTED]

The throwing power of the electro-less and electrolytic plating process is much higher than that of microvias. The fluid dynamics--the wetting of a trench --is much better than that of a microvia. This advantage is used to make the shielded construction smaller, enabling it to route between smaller pitch devices.

The (current) minimum width of a shielded and 50 [OMEGA] impedance controlled structure is 0.7 mm. To route into fine-pitch area array packages, a smaller structure (<0.5 mm) must be formed. Aspect ratios of blind via structures up to 10:1 are under study. FIGURE 3 shows a section of a high-aspect-ratio trench or "non-round" blind via in Thermount material.

[FIGURE 3 OMITTED]

Another aspect in drilling or trenching is data preparation and how to control the laser workstation. A simple "flat" drill file would work, but the number of drill positions will slow processing in CAM and at the laser. Setting position overlap requires reprocessing of the file (FIGURE 4). The preferred method is to use a vector-based format F where the overlap can be programmed and varied on the workstation, followed by a post-processing. Vector format results in smaller data packages, start and stop management around corners and crossovers (see Figure 4).

[FIGURE 4 OMITTED]

Design and Fanouts

Fanning out with a shielded trace is similar to using guard traces. The only difference is that one has to place guard traces on the layer above and below the shielded trace. Using this FIG technique in CAD systems, a full DRC can be executed without compromising the integrity of the design. (As the technology matures design tools will be equipped with features that apply automatically the shielding.) Current in the "trench" traces can be output with the rest of the design in Gerber of other format. At CAM the shielded traces will be selected and converted in an input format for the laser workstation. Any design system can be used to build shielded transmission lines.

Likewise, as is the case with conventional guard traces, the trenches can be routed around the termination point or via. In FIGURE 5 a SMD pad is shielded by trenches, minimizing the opening for radiation/crosstalk. The trenches are in this case directly connected to two ground pads (left and right from the signal pad). The width of the trench is narrowed when the signal is launched from the pad. Bringing the shielding closer to the signal will change the impedance.

[FIGURE 5 OMITTED]

Depending on the signal type, single or groups of traces can be shielded. A group of very sensitive signals that, for example, terminate in a coaxial cable can be routed in parallel under almost similar conditions for PCBs as for coaxial cable. FIGURE 6 shows sample individual shielded transmission lines. No crosstalk is possible between the shielded traces, making it a unique and perfect structure for high-speed or sensitive signals. In FIGURE 7, multiple traces are routed within a single shielding.

[FIGURES 6-7 OMITTED]

The three-layer shielded structure for a single layer or edge-coupled line and four-layer structure for broad side coupled lines can be placed on multiple layers in the build. Depending on the complexity of the structure multiple shield lines can be placed in the multilayer. FIGURE 8 shows a full section of a multilayer with shielded traces. In this build Thermount and FR-4 are combined.

[FIGURE 8 OMITTED]

The change in impedance by the shielding of the sidewall varies--as in any stripline--with the distance of the conductor to the shielding. FIGURE 9 shows the change in impedance as a function of the distance between the trench shielding and signal. The relationship is shown between the thickness of stripline and distance trench-signal. The consequence is that stripline models cannot be used to calculate impedance where the distance trench-signal is smaller than ~300 [micro]m (for ranges specified in Figure 9). (The plot in Figure 9 is calculated using Ansoft representing a rectangular representation of the structure.)

[FIGURE 9 OMITTED]

Similar to Microvias

How reliable is a plated trench? Let's first answer, how reliable is a microvia? The major failure mechanism for a microvia is barrel expansion, followed by mechanical damages in assembly/rework. The barrel length of a microvia is ~250[micro]m max., typical ~50 [micro]m. Assuming no process defects such as voiding or bounce pad cleanliness, microvias will outlast another element (e.g., a hole with longer barrel length). Same goes for the trench. The differences are that the depth of the trench will be typically ~400 [micro]m, with a potential maximum of ~550 [micro]m, and measuring a short (connection layer 1-3) over a huge contact area is difficult.

The value for the shielding quality is the transfer impedance. Measuring this value on embedded structures, which have multiple connections through several ground planes, needs to be resolved. Multiple techniques have been developed to measure coaxial cables, which is a "simple" element compared to boards.

Simulations have shown that voids as large as 75 [micro]m in the shielding (simulated @40 GHz) do not have a measurable effect on the performance; i.e., no crosstalk. Measuring voids using time domain reflectrometry (TDR) is possible, but a high sample rate is required to detect small deviations in the integrity of the shield. In practice this is not an option.

The other element to be researched is CAF (conductive anodic filament). The industry is conducting measurements to determine minimum distances for hole-to-hole or hole-to-trace as function of material and process. A similar test needs to be conducted for trench-to-trace. At the moment the distance between these elements is ~150 [micro]m the goal is to reduce this to 100 [micro]m.

A key enabler for the success of this technology is the aspect ratio of the trenches to fanout from finer-pitch devices. Bringing the shielding closer to the device will make the transmission line less sensitive for its environment. The next step is shielding the interconnection between the package and board. The use of shielded connectors, available from various sources, is a first step in making the interconnect over various devices insensitive for noise.

Spinoffs from the "shielding" technology include the formation of power traces. The goal is to fill/plate the trenches, similar to semiconductor manufacturing. The traces then can be used for high currents. The alternative technique is to use thick copper claddings. The disadvantage that no fine-line features can be routed on the same layer as the power traces. When using "vertically" oriented power traces, a double-sided PCB would be sufficient.

Note: The shielded transmission line technology and non-round via hole technology are patent pending.

Ed.: This article is adapted from a paper presented at EPC 2002 and is used with permission of the author.

ACKNOWLEDGMENTS

The author would like to acknowledge the Viasystems MicroCoax team, Dartan Broens, Martin Cotton, Paul Denomme, Patrick Lebens and Rob Wilms, for their input and effort in developing the technology as well as everyone who supported the work within Viasystems.

BIBLIOGRAPHY

Martin Cotton, "MicroFeatures and Embedded Coaxial Technology," Electronic Circuits World Convention, September 1999.

S. Secru, "High Speed Copper Links--Backplane and Cable Assemblies," DesignCon, January 2002.

JOAN TOURNE is advanced technology and business development director at Viasystems Mommers B.V. He can be reached at [e-mail].