Embedded Resistors: Squares, Trimming, and When It Pays
A buried resistor is just sheet resistance times a count of squares. The design is trivial; the economics and the tolerance are not. Here is what determines whether burying resistors in the stack-up is worth it.
Contents
The whole design is one division
An embedded resistor is a patch of resistive film laminated into the board. Its resistance is sheet resistance times the number of squares:
For a 100 Ω target on 25 Ω/sq nickel-phosphorus film at 1 mm width: 4 squares, so 4 mm long. That element occupies 4 mm² and handles about 100 mW.
Trimming, and why you design low
As-fabricated tolerance on embedded film is around ±10 %. That is fine for a pull-down and useless for anything setting a gain or a reference.
Laser trimming fixes it. You cut into the element, narrowing the current path, which raises resistance. Trimming can only go one direction, so the design must start below target:
- Design length 3.6 mm, giving 90 Ω pre-trim
- Trim up to the 100 Ω target
- Post-trim tolerance around ±5 %, and tighter with more trim time
The 10 % trim allowance in that example is the headroom. Too little and a low-side part cannot reach target; too much and you spend laser time on every board and take a bigger tolerance hit from the trim cut itself.
Where the value actually is
The case for embedded resistors is rarely cost per resistor. Discrete 0402s are extraordinarily cheap and the assembly line places them for a fraction of a cent.
The case is:
Density. Reclaiming surface area under a dense BGA, where there is no room for a termination network anywhere near the pins. Inductance. A buried resistor sitting directly in the layer stack has far less parasitic inductance than a 0402 with two solder joints and two fanout vias. For a series termination on a fast edge, that difference is the point — the termination is only as good as its parasitics. Reliability. No solder joints means no solder joint failures. For high-reliability and high-vibration hardware, removing hundreds of joints from the assembly is a meaningful reduction in failure modes. Count. A board with 200 identical terminations amortises the process cost in a way that a board with 12 never will.What it costs you
A laminate decision. Embedded resistor material is a specialty foil that must be in the stack-up from the start. You cannot add it later, and it constrains which fabricators can build the board. Panel yield. A failed embedded resistor is not reworkable. On a discrete design, a bad resistor is desoldered and replaced. On an embedded design, the panel is scrap. That risk is priced into every quote. No changes. A value change on a discrete resistor is a BOM edit. On an embedded resistor it is a new stack-up, new artwork, and new tooling. This alone rules it out for anything still in flux. Testing. The resistors are inaccessible after lamination, so verification depends on test coupons and the fabricator's process control rather than direct measurement of the part you shipped.Material choice
Nickel-phosphorus is the workhorse, available in a range of sheet resistances, with 25 Ω/sq and 100 Ω/sq being the common stock values. It trims well and has a reasonable temperature coefficient.
Tantalum nitride offers a much better temperature coefficient and long-term stability, at higher cost — the choice when the resistor is setting something precise over temperature.
Carbon and thick-film options reach much higher sheet resistances, which matters when a high value would otherwise need an impractical square count. A 100 kΩ resistor at 25 Ω/sq needs 4000 squares, which is not a resistor, it is a maze.
That square-count arithmetic is the practical constraint on value range. Embedded resistors live comfortably from tens of ohms to a few kilohms. Outside that, the geometry stops making sense.
Guidance
Only for stable designs. If the value might change, use a discrete. Check the square count early. If it is under about 0.5 or over about 100, pick a different sheet resistance rather than distorting the geometry. Watch power density. Buried film has no convection path — heat leaves only by conduction into the laminate. The area-based power limit is not conservative padding. Keep the aspect ratio sane. Very long thin elements are vulnerable to etch variation across their length; very short wide ones are dominated by end effects at the terminations.The embedded resistor calculator works square count, physical length, pre-trim design target, area, power limit, and trimmed and untrimmed tolerance for the common film materials.
Related Articles
Annular Ring: Why the Pad Is Bigger Than You Drew It
A 0.3 mm drill does not give you a 0.3 mm hole, and the pad around it has to absorb drill wander, registration error, and plating build-up. Here is where each of those numbers comes from and why IPC-6012 Class 3 costs more.
Aug 18, 2026
PCB DesignBGA Escape Routing Starts at the Land Pad
How many traces fit between two BGA balls is decided the moment you pick the pad diameter. Here is the arithmetic from pitch to channel width to trace count, and why NSMD and via-in-pad change the answer.
Aug 18, 2026
PCB DesignYour Stub Is 13% Too Short: εeff vs the Datasheet εr
A microstrip is half-buried in laminate and half in air, so the wave sees neither. Using the datasheet dielectric constant for a quarter-wave stub on FR-4 gets the length wrong by 13 percent — enough to move a filter edge hundreds of MHz.
Aug 18, 2026