The conclusion first: the layer assignment table is the only document in which the fab can read your layer-order intent, and what it reads there is not what you were thinking about when you drew it. On the design side you arrange the layer order around reference planes, return path length, and which two layers must not be adjacent. The fab receives the same table and looks at a different set of entries: which layers carry fine lines, which are essentially solid copper, where each layer's copper distribution will push resin during lamination, and which layer images become untouchable once pressed. Same table, different columns. The columns you left blank get assumed by the fab, and an assumed board is usually not the board you specified.
1. Design reads layer order, manufacturing reads "how each layer gets processed"
In a design review, layer assignment usually ends with one sentence: signal, ground, power, signal. That sentence carries no manufacturing information. The fab needs the same order translated into process steps: this layer is an inner image and goes to etch; that one is an outer layer and goes through pattern plating then etch; these two are the faces of one core; this gap is one sheet of prepreg and that one is three.
The translation is not automatic. The same sentence — "six layers, signal-ground-signal-signal-power-signal" — maps to at least two completely different manufacturing routes: three cores in a single lamination, or a four-layer press followed by added outer layers. The two differ in registration accumulation, resin demand, and whether buried vias are possible at all. Layer order does not specify the manufacturing route; the stackup table does. When the table carries only an order and no core/prepreg division, the fab either asks or arranges the route it runs most easily, and the interlayer structure of that board stops being the one you designed.
For the engineering premises behind layer count and structure selection, see the multilayer PCB page.
2. The entries the fab actually reads
Opened up, the table contains the following items the fab works through one by one. None of them is a signal integrity concept, which is exactly why they get treated in design review as "the fab will handle that."
First, whether each layer is fine-line or solid copper. This determines etch handling. A fine-line layer has dense, uniform-width features with one characteristic etch load and sidewall behaviour; a large solid plane has only clearances and isolation openings and loads the etchant completely differently. Marking which layer is which lets the fab separate them within a batch instead of running one parameter set against two very different image types.
Second, whether the gap between two adjacent layers is a core or prepreg. This decides whether the spacing is fixed by stock material (a core, whose thickness is already determined) or by the lamination process (prepreg, whose final thickness is known only after pressing). "These two layers need to be this thick" is not sufficient on its own — the same thickness realised through a core and through prepreg have different tolerance sources and different availability. For how stock constrains the structure, see core and prepreg availability.
Third, whether the copper on each layer is evenly distributed. This is the most frequently omitted item and it directly governs where resin flows during lamination. If half a layer is solid copper and half is open, resin is pushed toward the open half; if the facing layer happens to be arranged the same way, the two imbalances add. With no copper-distribution information in the table, the fab can only read it out of the Gerber, and what it can then do is limited — usually adding copper thieving, and whether the placement of that thieving gets confirmed with you depends on whether you left a place in the table for that conversation. For the resin flow parameter window itself, see press profile and resin flow.
Fourth, whether the order is symmetric about the mid-plane. That an asymmetric stackup bows after pressing is usually known on the design side, but the manufacturing criterion is finer: not only layer-count symmetry, but whether each opposing pair matches in copper weight and copper distribution. For how symmetry is read in manufacturing, see stackup symmetry requirements.
Fifth, which layers participate in which lamination cycle. On a sequentially laminated board this entry is the process order. It also sets which layer images must be made first and carry the thermal history of every subsequent press, and where registration accumulation begins. For the control measures, see interlayer registration control.
3. The three items designers omit most often
Ranked by how often they trigger a query or a hold, the omissions run roughly in this order.
Copper distribution is the most common. The table states copper weight per layer but says nothing about whether that copper is full or sparse. Copper weight is a material specification; copper distribution is an image fact. Both affect the lamination result, and only the first one made it into the table.
Second is "who has to realise this thickness." The design states an interlayer thickness but not whether prepreg stacking is acceptable to reach it. Some of those thicknesses came out of an impedance calculation and must hold; others are purely mechanical spacing and can be substituted with the nearest stock combination. Without that distinction the fab cannot tell where it may adjust and where nothing may move.
Third is the outer-versus-inner processing declaration. Outer layers go through pattern plating and gain copper on top of the base foil; inner layers do not. If the table gives a single "copper weight" without saying whether it is base or finished, that column has two readings. For material and copper engineering inputs, see the materials page.
4. What the fab does when something does not add up
There is a real split here worth knowing in advance: not every mismatch comes back as a question.
What comes back are the items where a fab decision would directly change electrical performance — impedance-driven interlayer thicknesses, dielectrics that must not be substituted, dimensions carrying an explicit tolerance. The fab will not settle those for you.
What does not come back are the items the fab considers to be inside its own process scope — how thieving is added, how many prepreg sheets make up an untoleranced thickness, which layers get grouped into one press. Those get arranged according to its standard practice and the board ships. Silence is not a signal that these do not matter; it is a signal that you did not indicate you cared. When a cross section later shows an interlayer structure other than the intended one, the responsibility boundary falls on the "not stated in the table" side.
So the table does more than convey information: it marks which items are not open to fab discretion. Items carrying a tolerance, a "no substitution" note, or a "thickness set by impedance" note are the ones actually locked. For the corresponding release items, see the stackup drawing release checklist.
5. How complete the table has to be before it can be released
The release criterion is not "detailed" but that whoever finishes reading it needs to make no assumptions. The practical check is three questions per layer:
What role does this layer have (fine line / solid plane / mixed)? What realises the gap between it and the layers above and below (core / prepreg / set by lamination), and may that gap be substituted? In which lamination cycle does this layer's image become fixed?
When every layer has a definite answer to all three, the table can go out. Any layer where any one question has to be answered with "the fab should know" is an open point, not a released input. For how to state tolerances, see the tolerance page; for the evidence types used in verification and release, see the quality page.
6. When to change the assignment instead of adding requirements
One case deserves separate mention: when the requirements on the table keep piling up — no substitution here, thickness locked there, and symmetry on top of that — sometimes the real problem is that the layer order itself is arranged too tightly, not that manufacturing capability is short.
Judge by the nature of the conflict. If it is "this thickness happens to fall away from common stock," that is an availability problem and a nearby layer order usually releases it. If it is "holding impedance breaks symmetry and holding symmetry moves impedance," that is a structural problem in the order; adding constraints to the table will not make it disappear, only defer it until after lamination. Changing the assignment is far cheaper at that point. Under HDI structures, layer assignment carries the additional dimension of lamination count — see the HDI PCB page and void and delamination prevention. For the overall capability scope, see the capabilities page.
FAQ
Q: Are the layer assignment table and the stackup drawing the same thing?
Not quite. The stackup drawing usually shows layer sequence and materials. The assignment table adds, on top of that, each layer's role, whether substitution is allowed, and which lamination cycle it belongs to. With sequence and materials only, the fab still has to supply the role judgement itself.
Q: Can the fab build from layer order alone, without a core division?
Yes, but then the fab picks the route. One order often supports more than one lamination scheme, and the schemes differ in registration accumulation and in which via structures are possible. To lock the structure, the division has to be in the table.
Q: Why state copper distribution when the Gerber already contains it?
The Gerber carries the fact; the table carries the intent. The fab can see from the Gerber that a layer is unbalanced, but not whether you accept thieving or which areas may receive it. The first is data, the second is authorisation.
Q: Does "copper weight x" in the table mean base or finished copper?
On outer layers the two differ; on inner layers they are essentially the same. Without a distinction the column has two readings. State explicitly which one applies to outer layers.
Q: What does the fab do with an untoleranced interlayer thickness?
It approaches the stated value with a common stock combination and accepts the result within its standard tolerance. If that thickness matters electrically, it has to be called out and marked as not substitutable.
Q: When symmetry and layer assignment conflict, which wins?
It depends where the conflict sits. First try a nearby set of interlayer thicknesses that preserves symmetry. If impedance and symmetry genuinely cannot both hold, that is a structural issue in the order and belongs back in the assignment, not in further constraints on the table.

