Most products land on four layers. Two layers is right for genuinely simple, low-speed circuits; six or more becomes necessary when you have dense high-speed routing, multiple power rails, or a part with a fine-pitch ball grid array. The layer count is not a cost-optimization detail to settle at the end of layout — it determines whether the board can meet its electromagnetic compatibility and signal integrity requirements at all, and changing it late means redoing the layout.
What the layers actually do
A multilayer board is copper layers separated by insulating material, connected by plated vias. Layers serve three purposes:
- Routing. More layers means more room for traces, which means shorter, more direct connections instead of long detours around obstacles.
- Reference planes. A solid, uninterrupted ground plane directly beneath a signal layer gives every trace a controlled impedance and a tight return current path. This is the reason four layers usually beats two, and it has little to do with running out of routing space.
- Power distribution. A plane instead of traces gives low-impedance power delivery, less voltage drop, and useful interplane capacitance.
The return current point deserves emphasis, because it is the least intuitive and the most consequential. Current returns to its source by the path of least impedance, which at high frequency means directly under the trace. Give it a solid plane and the loop area is tiny, so radiated emissions and susceptibility are low. Force it to detour around a split or a gap and you have built an antenna — which shows up as a failure at the test lab, not on your bench. That connection is why EMC testing outcomes are often decided by stackup choices made months earlier.
Rules of thumb by layer count
Two layers
Signals and power share both sides; a real ground plane is impossible because routing keeps cutting it. Appropriate for power supplies without fast switching, simple sensor boards, LED drivers, breakout boards, and low-frequency analog. Cheapest option, widely available, and entirely adequate when the circuit is genuinely slow.
Four layers
The workhorse: signal, ground, power, signal. Every trace has a plane close beneath it, decoupling works properly, and impedance is predictable. This is the right default for essentially any board with a microcontroller, a radio, USB, or a switching regulator. The step from two to four layers is the single highest-value stackup decision most projects make, and it typically costs far less than the engineering time spent fighting noise and impedance problems on a two-layer board.
Six layers
Adds a second routing layer pair between planes, or a second plane for a separate rail. Needed when a dense part cannot be escaped in four layers, when you have several supply voltages, when differential pairs need clean reference throughout, or when a sensitive analog section needs isolation from digital noise.
Eight and above
Application processors, DDR memory, high-speed serial links, and fine-pitch BGAs. At this level stackup becomes a specification you agree with the fabricator, including impedance targets, dielectric materials, and controlled build tolerances.
How layers drive cost
Bare-board price scales with layer count roughly stepwise, in pairs, since layers are added symmetrically. Going from two to four is a modest increase at prototype quantities — often tens of dollars on a small board, and much less per unit at production volume. Going from four to six or eight increases both the layer charge and the process risk.
Layer count is rarely the dominant cost driver, though. What actually moves board price:
- Board area — usually the biggest factor, since panels are priced by usable area.
- Minimum trace width and spacing — below common thresholds, yield drops and price rises.
- Via technology — through-holes are standard; blind, buried, and microvias add significant cost and process steps.
- Material — standard laminate versus a controlled-loss high-frequency material.
- Quantity and lead time — expedited small runs pay a large premium, which is the main variable behind PCB prototype cost.
The false economy is choosing two layers to save a little on the board and then paying for extra spins, filtering components, shielding, and a failed compliance test. Board cost is a small fraction of total product cost; a redesign is not.
Considerations beyond price
- Thickness and rigidity. More layers generally means a thicker, stiffer board. That interacts with connector footprints, card guides, and enclosure fit.
- Thermal behavior. Internal copper planes spread heat effectively, which is often the cheapest thermal fix available — see thermal management in electronic products.
- Antenna behavior. Ground plane geometry is part of the antenna. A stackup change shifts tuning, so coordinate it with antenna design for wireless products.
- Protection circuitry. Clean low-inductance paths to ground are what make suppression devices work; they need plane access, as discussed in ESD protection circuit design.
- Flexible sections. If part of the circuit must bend, layer count trades directly against flexibility — see when to use a flex PCB.
- Assembly and rework. Heavy internal copper sinks heat during soldering and makes hand rework harder.
How to decide, in order
- List the fast signals: clock rates, switching edges, radio frequencies, differential pairs. Anything fast means at least four layers.
- Count supply rails. Three or more usually pushes toward six.
- Check the densest package. Look at the escape routing a fine-pitch BGA needs, since that alone can set the count.
- Confirm board outline and mechanical constraints, because area and thickness are fixed by the enclosure.
- Agree the stackup with your fabricator before layout begins, and get impedance targets confirmed rather than assumed.
Settle this during architecture, not during routing. Stackup changes late in the process invalidate impedance work and often force a re-spin, which pushes into the build phases described in EVT, DVT, and PVT.
How Projects House approaches it
We choose the stackup from the signal list and the compliance target before a single trace is drawn, because layer count is a requirement rather than a preference. More on this area is collected on our electronics development page.
Not sure what your board needs?
Send us your schematic or requirements through the contact form and we will recommend a stackup that meets your signal, thermal, and compliance needs without paying for layers you do not use.