How Do You Fan Out and Escape-Route a Fine-Pitch BGA?
Last updated 15 July 2026 · 5 min read
Direct Answer
Fanout (also called escape routing) is the process of connecting every ball on a BGA package to a via that brings the signal to a routing layer where it can travel freely — it happens before general board routing begins, because a BGA's balls are only accessible from directly underneath the package, and every ball not on the outermost rows is physically blocked from a direct trace connection by the rows of balls surrounding it. The standard technique is a dogbone pattern: a short trace runs from each ball pad outward to a via placed just outside the ball grid, named for the bone-like shape the pad-trace-via combination makes. How many balls can actually be escaped, and with what via technology, is set almost entirely by the package's ball pitch — a coarse 1.0 mm pitch BGA can often be escaped with ordinary through-hole vias and standard trace/space rules, while a fine 0.4-0.5 mm pitch package typically forces microvias or via-in-pad just to physically fit a via between adjacent balls at all.
Detailed Explanation
A BGA (Ball Grid Array) package presents every one of its connections as a solder ball on a grid under the package body — unlike a QFP or SOIC, there's no lead sticking out from the package edge that a trace can simply run to. Only the balls on the outer edge of the grid have a clear path to the surrounding board; every ball inside that perimeter is physically surrounded by other balls on all sides, and the only way to reach it is straight down, through the board, via a via placed at or immediately next to that specific ball. Fanout (or escape routing) is the process of giving every ball that via, so the signal reaches a layer where normal routing can take over — it's a distinct, earlier step from general board routing, and it has to happen first, because the fanout pattern determines how much (or how little) routing channel is left between vias for anything else that needs to cross under the package.
The standard fanout pattern is the dogbone: a short trace runs from the ball's own SMT pad outward to a via placed just clear of the ball grid, rather than putting the via directly in the pad. The pad-trace-via shape resembles a dog's bone, which is where the name comes from. This is the default because it keeps the via (and its annular ring) off the solder pad itself, avoiding the wicking and voiding risk a via directly under a solder joint can create during reflow — see via-in-pad for the alternative where the via genuinely does go inside the pad, and why that requires an added fill-and-cap fabrication step to be reflow-safe.
Practical Examples
A 1.0 mm pitch BGA (common on many microcontrollers and simpler ICs) is usually the easiest case: the pitch is generous enough that ordinary through-hole vias fit comfortably in a dogbone pattern on a standard trace/space capability, and even boards with only four or six layers can often escape the whole package without needing microvias or via-in-pad at all.
A 0.5 mm (or finer) pitch BGA — typical of a high pin-count FPGA, SoC, or DDR memory device — is a fundamentally different problem: at that pitch, there is often not enough physical space between adjacent balls to fit a standard-size through-hole via and its annular ring at all, which is exactly the scenario that drives the microvia and via-in-pad discussion in types of PCB vias. The layer count and stack-up itself are frequently set by this requirement before the rest of the board's routing is even considered — see what is a PCB stack-up? for how BGA escape routing drives the minimum layer count on a dense design.
Design Considerations
- Calculate routing channel capacity from the actual ball pitch and your fab's real trace/space capability, not from a number that worked on a previous project's different-pitch package — see the FAQ above for how this changes with pitch.
- Route the outermost 1–2 rows of balls directly where possible, saving via count and routing layers for the interior rows that have no alternative but to escape through a via.
- Plan via technology (through-hole, microvia, or via-in-pad) against the specific package's pitch before committing to a stack-up — a package pitch that's genuinely too fine for through-hole vias will force a stack-up and fabrication cost decision that needs to be made early, not discovered after layout has already started.
- Stagger interior vias where the pitch allows it, rather than placing every via on a strict rectangular grid — a staggered pattern can open up marginally more routing channel between vias on a tight-pitch package, though the achievable pattern is ultimately set by the ball grid itself.
- Verify escape routing against your fab's actual annular ring, minimum via size, and minimum trace/space capability before finalising the pattern, since a fanout that looks fine in the layout tool can be unmanufacturable if it assumes a tighter tolerance than the chosen fab can reliably produce.
- Fine-pitch BGA layout: escaping a dense, fine-pitch package correctly the first time avoids a costly stack-up or layer-count respin partway through a project — professional PCB layout services plan the fanout and via strategy against the fab's actual capability as part of the initial stack-up decision.
Common Mistakes
- Starting general board routing before the BGA fanout is fully planned, then discovering partway through that the escape pattern doesn't leave enough routing channel for the rest of the board's nets to pass underneath the package.
- Assuming a routing-channel width or via pattern that worked on a previous, coarser-pitch package will carry over to a finer-pitch one without recalculating against the new pitch and the fab's actual capability.
- Placing interior-row vias without first confirming the fab can actually manufacture a via and its annular ring at that pitch, discovering the conflict only when the design comes back from a manufacturability check or fails at the fab house.
- Treating dogbone vias and via-in-pad as interchangeable defaults rather than a deliberate choice driven by pitch — using via-in-pad everywhere adds unnecessary fill-and-cap fabrication cost on rows where a standard dogbone pattern would have fit comfortably.
Frequently Asked Questions
- How many traces can actually be routed between two adjacent BGA vias?
- It depends entirely on the fab house's minimum trace width and clearance capability relative to the via pitch, not on a fixed industry number. For a 1.0 mm pitch BGA with vias placed on that same 1.0 mm grid, a fab capable of 4 mil (0.1 mm) trace and 4 mil clearance can typically fit one trace between two adjacent vias with room to spare; a 0.5 mm pitch package on the same fab capability leaves far less room, often forcing a single trace per channel at best, or requiring a finer trace/space capability to fit even that. Always calculate this from the actual pitch and the actual fab's manufacturing capability for the specific board — don't assume a routing-channel width that worked on a previous, differently-pitched package will carry over.
- Does every ball need its own via, or can some route out directly?
- Only the outermost row (or two, depending on pitch and via technology) of balls can realistically route out directly with a short trace to the board edge of the footprint without needing a via at all. Every ball inside that outer perimeter is physically enclosed by surrounding balls and has no path to general routing except straight down through a via placed at or near that ball's own location — which is exactly what fanout/escape routing exists to provide.
References
Related Questions
What Are the Different Types of PCB Vias, and When?
PCB vias connect copper across layers. Here's how through-hole, blind, buried, and microvias differ, and how to choose the right type for a layout.
What Is a PCB Stack-Up, and How Do You Design One?
A PCB stack-up is the arrangement of copper and dielectric layers in a board. How to choose layer count, copper weight, and laminate material.
What Is PCB Routing, and How Do You Route a Board Well?
PCB routing draws the copper traces implementing a schematic's netlist, balancing trace width, spacing, layer use, and impedance requirements.
PCB Component Placement: Best Practices and Common Pitfalls
Good PCB component placement groups related parts, shortens critical paths, and separates noisy from sensitive circuitry. Here's how to do it well.
What Is Controlled Impedance PCB Design, and Why It Matters?
Controlled impedance PCB design shapes trace geometry so a trace presents a specific, predictable impedance. Here's when you need it and how.
PCB Design for Manufacturability (DFM): What It Means
PCB design for manufacturability (DFM) matches a layout to a fab and assembly house's real process capability, preventing costly late-stage respins.
Related Forum Discussions
Is a double-sided PCB enough for a simple ESP32 sensor board, or should I go multi-layer?
Building a little battery-powered sensor board around an ESP32 module (the kind with the PCB antenna already built into the module, not desi
Decoupling caps placed right next to the IC but still seeing power rail noise
Fast digital board, MCU running a few hundred MHz core clock plus a couple of high-speed peripherals. Put 100 nF caps right next to every po
Best 4-layer stack-up for a board with both analog and digital circuitry?
Spinning up a board with an STM32 doing the digital side and a precision analog front end (instrumentation amp into an ADC) reading a low-le