Resistor Tolerance and Network Stacking

What a Tolerance Rating Actually Bounds

Every resistor is marked with a nominal value — 10 kΩ, 4.7 kΩ, 220 Ω — but that number is never the exact resistance sitting on the board. Manufacturing variation means the true resistance lands somewhere in a band around the nominal value, and the tolerance rating printed or color-coded on the part (±10%, ±5%, ±1%, and tighter) is the guaranteed width of that band. A resistor marked “10 kΩ ±5%” is not promising exactly 10,000 Ω — it is promising a real value anywhere from 9.5 kΩ to 10.5 kΩ, with no further indication of where inside that range any individual part actually sits.

This distinction matters because it changes what a “10 kΩ resistor” means in a design. If a circuit needs a resistance within 1% of some target, a single ±5% resistor cannot deliver that on its own no matter how precisely you read the schematic — the part itself does not hold that precision. Tolerance is a property of the component, not of your measurement or your calculation, and it sets a hard floor on how accurately any circuit built from that part can be characterized.

The minimum and maximum bounds are simple to compute for a single resistor: multiply the nominal value by (1 − tolerance) for the minimum and by (1 + tolerance) for the maximum. The harder question — and the one that actually matters for real designs — is what happens to that band once you combine two or more resistors into a network.

Series Networks: Worst-Case Bounds Add Directly

Series resistance is a plain sum: R_total = R1 + R2 + … + Rn. Because the total is a linear, monotonically increasing function of every individual resistor, increasing any one resistor can only increase the total, and decreasing any one resistor can only decrease it. That means the worst-case minimum total occurs when every resistor in the chain sits simultaneously at its own minimum, and the worst-case maximum total occurs when every resistor sits simultaneously at its own maximum.

In practice this is just addition: sum each resistor's minimum bound to get the series minimum, and sum each resistor's maximum bound to get the series maximum. No special formula is needed beyond the one you already use for nominal series resistance — you simply run it twice, once on the low bounds and once on the high bounds.

When every resistor in the chain shares the same tolerance percentage, the resulting percentage tolerance on the total stays close to that same percentage — the absolute error in ohms scales with the total resistance in almost the same proportion. It is only an exact match when every resistor shares an identical value and tolerance; mix different values or different tolerance grades and the combined percentage shifts slightly, which is exactly why it is safer to work the actual min/max sums rather than trying to average percentages by eye.

Parallel Networks: Why Percentages Don't Simply Add

Parallel resistance is not linear: 1 / R_total = 1 / R1 + 1 / R2 + … + 1 / Rn. You cannot take each resistor's tolerance percentage and combine them arithmetically the way you can in a series chain — the reciprocal relationship means a given percentage change in one resistor does not translate into the same percentage change in the parallel total. Adding percentages here would give you a number that looks plausible but is not the actual worst-case bound.

What does hold is monotonicity: increasing any single resistor in a parallel group always increases the overall parallel resistance, and decreasing any single resistor always decreases it. That is enough to locate the true extremes. The worst-case minimum parallel resistance occurs when every resistor in the group is simultaneously at its own minimum value, and the worst-case maximum parallel resistance occurs when every resistor is simultaneously at its own maximum value. You do not mix a minimum on one branch with a maximum on another — the extremes only occur when every resistor moves the same direction at once.

The reliable method is therefore to evaluate the parallel formula twice: once feeding in every resistor's minimum bound, and once feeding in every resistor's maximum bound. Whatever topology the network uses, this “all-low, then all-high” evaluation is what produces the correct worst-case envelope — not a shortcut built from adding or averaging percentages.

Worked Example: Two 10 kΩ ±10% Resistors, Series vs Parallel

Take two identical resistors, each marked 10 kΩ ±10%. Each individual part has a minimum of 9 kΩ (10 kΩ × 0.9) and a maximum of 11 kΩ (10 kΩ × 1.1).

In series: the nominal total is 10 kΩ + 10 kΩ = 20 kΩ. The worst-case minimum is the sum of both minimums, 9 kΩ + 9 kΩ = 18 kΩ, and the worst-case maximum is the sum of both maximums, 11 kΩ + 11 kΩ = 22 kΩ. The series pair lands anywhere from 18 kΩ to 22 kΩ around a 20 kΩ nominal — a band that, worked out as a percentage of 20 kΩ, is still ±10%.

In parallel: the nominal total is (10 kΩ × 10 kΩ) / (10 kΩ + 10 kΩ) = 5 kΩ. Running both resistors at their minimum gives (9 kΩ × 9 kΩ) / (9 kΩ + 9 kΩ) = 4.5 kΩ, and running both at their maximum gives (11 kΩ × 11 kΩ) / (11 kΩ + 11 kΩ) = 5.5 kΩ. The parallel pair lands anywhere from 4.5 kΩ to 5.5 kΩ around a 5 kΩ nominal.

Notice that both topologies happen to produce a ±10% band around their respective nominal values here — but that is a feature of using two identical resistors with the same tolerance, not a rule you can rely on. Swap in a second resistor with a different value or a tighter tolerance grade, and the parallel result's percentage band will no longer match either input's individual tolerance. The only way to know the real bound in that case is to run the min/max evaluation on the actual values involved, exactly as shown above.

Mixed Networks and the General Rule

Real circuits often combine series and parallel stages — two resistors in parallel feeding into a third in series, or a series pair forming one branch of a larger parallel group. The good news is that the same principle scales to any topology, no matter how many stages it has: evaluate the network's resistance formula once with every resistor pinned to its own minimum bound, and once with every resistor pinned to its own maximum bound. The two results are the network's true worst-case minimum and maximum.

This works because every series and parallel combination formula is monotonically increasing in each of its resistor inputs — larger component values never decrease the equivalent resistance of a series or parallel group. That property carries through function composition, so it holds for a mixed network built by nesting series and parallel stages together, not just for a single flat stage of either kind.

It is worth being clear about what this worst-case bound represents: it is the guaranteed envelope assuming every resistor in the network happens to land at the same extreme of its tolerance simultaneously — the least statistically likely outcome, not the typical one. Real batches usually cluster closer to nominal, and errors in different parts often partially cancel rather than reinforcing each other. Worst-case bounds exist to give you a guaranteed design margin, not a forecast of what a given circuit will actually measure.

Explore the Tool

Related Guides

Frequently Asked Questions

What does a resistor tolerance of ±5% actually mean?

It means the manufacturer guarantees the resistor's real resistance falls within 5% of the marked (nominal) value, in either direction. A 10 kΩ ±5% resistor could measure anywhere from 9.5 kΩ to 10.5 kΩ and still be within spec — the printed value is a center point, not a promise.

Can I just add the tolerance percentages together for resistors in series?

Not directly as percentages, but the underlying math is still simple addition. Sum each resistor's minimum resistance for the series worst-case minimum, and sum each resistor's maximum for the worst-case maximum. The resulting percentage tolerance on the total is usually close to, but not exactly, the individual percentage, because the absolute error amounts add while the nominal base also changes.

Why can't I just add tolerance percentages for resistors in parallel?

Parallel resistance is a nonlinear function of its inputs, so percentage errors do not pass through unchanged. The correct approach is to recompute the parallel formula twice — once with every resistor at its minimum value and once with every resistor at its maximum value — because parallel resistance moves in the same direction as its inputs: all resistors low gives the lowest possible parallel resistance, and all resistors high gives the highest.

Does worst-case tolerance stacking apply to mixed series-parallel networks too?

Yes. The same principle extends to any topology: evaluate the whole network once with every resistor pinned to its minimum tolerance bound, and once with every resistor pinned to its maximum bound. The two results are the true worst-case minimum and maximum for the network, regardless of how many series and parallel stages it has.

Is the worst-case bound the same as the statistically likely error?

No. The worst-case bound assumes every resistor in the network happens to land at the same extreme simultaneously, which is the least likely outcome in practice — most real batches cluster closer to nominal, and errors from different parts tend to partially cancel. Worst-case bounds are a guaranteed envelope for design margin, not a prediction of typical measured error.

Do Resistora's calculators handle tolerance stacking automatically?

Yes. The Circuit Calculator and Series/Parallel Calculator both have a Tolerance option — switch it on and enter each resistor's tolerance, and the tool computes the worst-case minimum and maximum resistance of the whole network for you, using the same min/max evaluation described above.