Multilayer PCB Advantages in High Speed Design

Once a design contains signals whose edges are fast relative to the length of the traces they travel along, the board stops being a container for wires and becomes part of the circuit. At that point the number of layers stops being a cost decision and becomes an electrical one. The multilayer PCB earns its place because it lets the designer control two things that a two-layer board cannot: the reference that every signal returns through, and the impedance of the distribution network that supplies it.

Signal Integrity

The primary advantage is signal integrity, and it comes from the availability of routing layers. With more layers, a trace can be routed on a layer where a plane lies directly beneath it, which gives the signal a predictable impedance and a return path that follows it closely. Short, regular routes with a consistent reference have lower loss, less crosstalk and far less reflection than the winding routes a two-layer board forces.

Impedance control is the mechanism. On a multilayer board the dielectric height between a trace and its reference is known, so the width needed for a 50 ohm trace can be calculated and held. The impedance discontinuities that occur at corners, via transitions and layer changes are then small relative to the characteristic impedance, and the reflections they produce remain within the timing margin.

The routing also becomes more regular. Because the designer is not fighting for space on two surfaces, traces can be planned as a group rather than routed one at a time, and the length matching required by parallel buses becomes achievable. The behaviour of those buses is described in this article on high frequency traces and data buses.

Power Integrity

The second advantage is power integrity. A multilayer board can dedicate one or more layers to power distribution and place them adjacent to ground planes, forming a distributed capacitor between the two. That capacitance supplies the high-frequency current that devices demand during switching, and it does so without the inductance that a trace would add.

The lower impedance of the distribution network has two visible effects. The supply voltage at the device remains closer to its nominal value during transients, which improves timing margin, and the loop area of the current drawn by each device is reduced, which reduces emissions. Where the power layer must be divided between rails, the geometry of the division matters, and the rules are set out in this discussion of power plane splitting.

Ground planes also provide the return path for signals, and more ground layers mean more opportunities to keep that return continuous. This is what makes the difference between a design that passes emissions testing and one that does not, since a broken return path converts a quiet trace into an antenna.

Electromagnetic Compatibility

The third advantage is electromagnetic compatibility, and it follows from the first two. Closely spaced layers contain the fields associated with fast signals, so less energy escapes the board and less external energy reaches the signals. Traces buried between planes radiate almost nothing from the board surface, because the field is confined within the dielectric.

The controlled structure also reduces the emissions caused by the power distribution network. A closely spaced plane pair does not radiate efficiently from its edges, and the stitching vias that tie the planes together remove the resonances that would otherwise amplify emissions at particular frequencies. The arrangement of layers that achieves this is discussed in this article on EMI reduction through stackup and layout.

EMC and the Cost Equation

`r`n

The electromagnetic compatibility benefit is what makes the additional layers affordable in practice. A design that fails an emissions test has to be corrected, and the corrections available after the layout is released are limited to filters, shields and cable treatments, all of which add cost and volume to the product. Controlling emissions in the stack removes that risk before it exists, and the rules that govern the arrangement are set out in this discussion of EMI reduction through stackup and layout.

`r`n

It is worth being honest about the cost. Each additional layer increases the board price, and the increase is not small. The justification is not that more layers are better in principle, but that the cost of the layers is lower than the cost of the alternatives: a design that needs six layers and is built on four will require additional components, a slower interface or a larger enclosure. Where the signal speeds are moderate and the reference arrangements are good, four layers often suffice, and the decision should follow from the requirements rather than from habit.

`r`n

Practical Implications for Layout

Layer count should be chosen before placement, because it determines the routing resources available. A four-layer board is the minimum for a design with fast signals, arranged so that both signal layers have an adjacent plane. Six layers allow a dedicated ground beneath each signal layer and a split power plane for analog and digital rails. Eight layers and above allow the fastest nets to be routed as stripline between ground planes.

Not every net needs the same treatment. Slow control signals can share outer layers, while clocks and high-speed links belong on the shielded inner layers. The rules that govern the layer assignment, and the impedance targets that follow from it, are described in this discussion of PCB routing techniques.

Layer assignment is also where the reference discipline is enforced. A signal that changes layers must change its reference at the same point, which means a ground via beside every signal via on a fast net, and a plane that is continuous beneath the whole route. Those requirements are simple to state and easy to overlook in a dense layout, which is why they belong on the review checklist rather than in the designer’s memory.

`r`n

Once the stack is fixed, the designer should confirm that every signal layer has a reference, that the reference beneath each fast net is continuous, and that the ground planes are stitched with a regular via pattern. Those three checks capture most of the benefit that the additional layers were purchased to provide, and they cost nothing beyond the review time required to walk the artwork once with the planes visible.

FAQ

How many layers does a high-speed design need? Four is the practical minimum, arranged so that each signal layer has an adjacent plane. Six layers allow separate references for analog and digital sections, and eight or more allow the fastest nets to be routed as stripline, fully enclosed between ground planes.

Does a multilayer board always perform better than a two-layer board? Only if the planes are used properly. A multilayer board whose signal layers have no adjacent reference performs no better than a two-layer board for signal integrity, and it costs more. The benefit comes from the reference arrangement, not from the layer count itself.

How does a multilayer board improve power integrity? By allowing a power plane and a ground plane to be placed close together, which creates a distributed capacitance across the whole board and lowers the impedance of the distribution network. That supplies transient current to devices without the inductance a trace would introduce.