Choosing a High-Speed Laminate for Server Boards

A server board is expected to run continuously for years at an elevated temperature, inside a chassis where airflow is the only cooling and where a failure means a service call. That combination puts demands on the laminate that a consumer board never faces, and it means the material has to be chosen for reliability as much as for loss.

This article covers the properties that matter when selecting a high-speed laminate for server and data centre equipment, and how to decide whether a premium material is justified.

What Server Boards Demand From a Laminate

The first demand is electrical: the channel between a processor and a memory device or a link partner has to close within its loss budget. The second is thermal: the board has to tolerate a sustained operating temperature that pushes an ordinary laminate close to its limit. The third is reliability over time, including resistance to the failure mechanisms that only appear after thousands of hours at temperature and voltage.

Those three demands pull in different directions. A material chosen purely for the lowest loss may have a lower glass transition temperature than a cheaper one, and a material chosen for thermal robustness may not be the lowest-loss option. The selection is therefore a compromise that has to be documented rather than a single figure read from a datasheet.

High speed laminate stack for a server motherboard

Loss Tangent and Channel Reach

The loss tangent sets the dielectric contribution to attenuation, and its effect grows with frequency and with trace length. For a short link inside a chassis almost any laminate is adequate, while a long channel between boards may only close with a low-loss material.

The conductor contribution has to be considered at the same time. In a low-loss material the dielectric term falls far enough that the copper surface roughness becomes the larger contributor, at which point a smoother foil returns more than a lower loss tangent would. The calculation should therefore compare the two terms rather than optimising one of them. The low-loss laminate selection notes describe how the choice follows from a loss budget, and the comparison of FR-4 and high-frequency stacks covers the stack implications.

Coupon test structures on a server backplane panel

Glass Transition Temperature and Thermal Ageing

The glass transition temperature is the point at which the resin changes from a rigid to a softer state, and a laminate operated near that point behaves differently from one operated well below it. The coefficient of thermal expansion increases sharply above the transition, which raises the strain on the plated barrels of every via at each thermal cycle.

Server boards are therefore built on materials with a transition temperature well above the operating range, typically well over 170 degrees Celsius, and the long-term thermal ageing behaviour is considered alongside it. A laminate that retains its mechanical properties after a thousand hours at temperature is worth more than one whose initial figures look better and degrade faster.

CAF Resistance and Long-Term Reliability

Conductive anodic filament growth is a failure mechanism in which a conductive path grows through the laminate between two conductors under the combined influence of voltage, moisture and temperature. It is a slow process that appears after long service, and it is one of the reasons server equipment specifies laminates with good resistance to it.

CAF resistance depends on the resin chemistry, the glass weave and the quality of the lamination, and it is improved by adequate spacing between vias and traces in the stack. That is a design measure as well as a material one, and it is one of the reasons the trace width and current calculation and the associated spacing rules are applied more conservatively on equipment that has to last for years.

Dimensional Stability and Registration

A material that moves during processing cannot hold tight registration between layers, and on a board with many layers the accumulated error becomes significant. Dimensional stability is quoted as a percentage change through the process, and a low figure allows the artwork to be scaled once and used without adjustment.

Stability also matters for impedance. The dielectric thickness is part of the impedance calculation, so a laminate that changes thickness during lamination produces a board whose impedance differs from the design value. High-performance materials are characterised for this behaviour, and the figures should be requested from the supplier rather than assumed.

Process Compatibility

A laminate has to be processable by the fabricator who will build the board. The required drilling parameters, the lamination cycle, the surface preparation before plating and the achievable hole quality all differ between material families, and a fabricator who does not build the material routinely will produce a board that meets the drawing and misses the electrical performance.

That makes the supplier relationship part of the material decision. Where the material is new to the supplier, a first article with coupons is essential, and the coupon data should be reviewed before the production order is placed. Impedance coupons, via chains and thermal stress results from the actual process are the evidence that the material has been handled correctly.

Cost and Where It Pays

A high-performance laminate can cost several times more than FR-4, and on a large multi-layer board the material can become the dominant cost of the bare board. The justification comes from the channel budget, the thermal requirement or the reliability requirement, and it should be stated in those terms.

Before changing the material, the cheaper options should be exhausted. Shortening the channel, improving the stack arrangement so that the high-speed layers are closer to their reference planes, and reducing the number of vias in the path all recover margin at no material cost. Only when those options have been considered does the laminate change become the right answer.

Moisture Absorption and Handling Before Assembly

A high-speed laminate absorbs moisture from the air like any other resin system, and the absorbed water changes both the dielectric constant and the behaviour of the material during soldering. The shift in dielectric constant is small but not always negligible on a tight impedance tolerance, and the more serious effect appears at reflow, where trapped moisture turns to steam and expands faster than the resin can relieve it.

The result is delamination, blistering between layers, or a barrel crack in a via, and it may not be visible from the surface. Server boards often pass through several assembly cycles, so the risk accumulates rather than being a single event. The material datasheet quotes a moisture absorption figure, and a low figure is one of the quieter advantages of a high-performance laminate over an inexpensive one.

Baking, Storage and the Effect on Solderability

The standard answer is to bake the panels or the assembled boards before the soldering step, following the profile the material supplier recommends. The profile matters: a bake hot enough to drive off moisture quickly can oxidise the copper surface and leave it difficult to wet, which trades one defect for another. A longer bake at a lower temperature is usually the safer route, and the supplier guidance should be followed rather than a workshop habit.

Storage is the other half of the problem. Panels held in an uncontrolled space will reabsorb moisture within days, so a board that was baked and then left on a shelf has to be treated as unbaked. Sealed bags with desiccant, a humidity indicator card and a defined floor life are the ordinary controls, and on long-life equipment they are worth writing into the assembly documentation. Thermal ageing and moisture damage interact as well, because a laminate already stressed by heat has less reserve when the next reflow arrives, so the two effects should be assessed together rather than as separate tests.

FAQ

Is a low-loss laminate always better for a server board? Not necessarily. A material with excellent electrical figures but a low glass transition temperature may be unsuitable for continuous operation at high ambient, so the thermal and reliability requirements have to be considered together with the loss.

How is the operating temperature of a laminate determined? From the ambient inside the chassis plus the temperature rise caused by the board’s own dissipation, with a margin for the worst-case airflow condition. That figure, rather than the room temperature, is what the material has to tolerate.

What evidence should be requested from the fabricator? Coupon data from the production process: impedance measurements, via chain continuity and resistance after thermal stress, and a cross-section confirming the dielectric thicknesses. Those results demonstrate how the material behaved, not what the datasheet claims.

Leave A Comment