Verbatim Article 100 "Continuous Load" (3+ hours), NEC 210.19(A)(1)(a) and (b), 210.19(A)(2), 210.20(A), the Table 210.21(B)(2) maxima, and the 210.19(A) 3%/5% informational notes — with eight worked examples where every number is computed by the shipped calculator core and asserted in the public test suite. Last updated 2026-08-31 (written for PanelWright v1.16).
Runs in your browser. No account, no install, data never leaves your machine. The panel card sizes every branch circuit against its load, the 220.82 card sizes the service, the Conductor derating card applies the 310.15 corrections that 210.19(A)(1)(b) demands, and the voltage-drop card checks the 3%/5% informational notes — the exact pipeline this article walks through.
A continuous load is one whose maximum current is expected to continue for 3 hours or more (Article 100). Where a branch circuit supplies such a load — alone or mixed with noncontinuous loads — the conductor ampacity and the overcurrent device rating must each be at least the noncontinuous load plus 125% of the continuous load (210.19(A)(1)(a) and 210.20(A) respectively) — and on a multi-receptacle circuit the ampacity must also meet the circuit rating (210.19(A)(2)), with the derated (310.15-corrected) ampacity still covering the load (210.19(A)(1)(b)). The same formula is why a 20 A circuit carries at most 16 A continuous: 20 ÷ 1.25 = 16, i.e. 80% of the rating (Table 210.21(B)(2) encodes the same number as receptacle load maxima).
Two things about this definition matter. First, it is about the maximum current — a load runs hot at its rating, not at its average. Second, "expected to continue" is a design judgment, not a usage log: a dedicated space-heating circuit in a dwelling is a continuous load; a 15 A branch circuit feeding a 12 A space heater that runs all night is one too. The 3-hour threshold is why the code multiplies by 125% — conductors and terminations driven at full current for 3+ hours settle at a higher steady-state temperature than anything duty-cycled, and the extra 25% is the thermal headroom for that steady state.
Read (a) and (b) together, because most real-world mistakes are one or the other applied alone. (a) sets the base ampacity from Table 310.16: noncontinuous + 125% of continuous. (b) then says the derated ampacity — base × 310.15(B)(1) ambient factor × 310.15(C)(1) conductor-count factor — must still carry the maximum load. A cold, empty conduit at 30 °C with three current-carrying conductors makes (b) a non-issue (both factors 1.00); a hot attic raceway full of 8 CCC makes (b) the controlling constraint, and the wire has to grow even though (a)'s number didn't change. The Table 310.16 article and the conductor-sizing end-to-end article cover the factor tables in full; this article's EX7 works a crowded-raceway case through the shipped core.
This is the clause people skip: even if (a)'s number is small, a circuit with more than one receptacle must have conductors rated for the whole circuit rating. A 30 A circuit carrying only 21 A noncontinuous + 5 A continuous needs (a) = 27.25 A of ampacity — but (A)(2) says 30 A minimum. It is also why 210.18 lets the breaker set the circuit rating ("Where conductors of higher ampacity are used for any reason, the ampere rating or setting of the specified overcurrent device shall determine the circuit rating"): over-sized conductors are legal for strength or voltage drop, but the circuit is only as big as its OCPD, and (A)(2) pegs the wire to that rating.
The formula is identical to 210.19(A)(1)(a) — that is not an accident. The code sizes the wire for steady-state heating (ampacity) and the breaker for the same steady-state current (rating), so the two agree by construction. Where they part company is at the 240.4(D) small-conductor caps: the breaker on 14 AWG copper may be 15 A maximum, on 12 AWG 20 A maximum, on 10 AWG 30 A maximum — regardless of what the 75 °C column of Table 310.16 says about the wire. EX5 works the case where the cap forces a size up. The full 240.4/D treatment lives in the conductor-sizing end-to-end article.
| Table 210.21(B)(2) — Maximum cord-and-plug-connected load to receptacle (A) | |||
|---|---|---|---|
| Circuit rating (A) | 15 or 20 | 30 | |
| Receptacle rating (A) | 15 | 20 | 30 |
| Maximum load (A) | 12 | 16 | 24 |
12 = 15 × 0.80, 16 = 20 × 0.80, 24 = 30 × 0.80. The SunCam 2023 NEC PDH calls out the symmetry explicitly: "this is 80% of the possible loading — which is a normal design criteria. The 80% (0.8) value also correlates with a 125% loading calculation for overload protection … because 100% divided by 80% equals 125%, or 1/0.8 = 1.25." So when you size a receptacle circuit, the 125% multiplier and the 80% table entry are the same constraint: the device must trip (and the wire must be sized) for 125% of the continuous load, which caps the continuous load at 80% of the device rating.
This is the only place in 210.19 where numbers are recommended rather than required. 3% on the branch, 5% total feeder-plus-branch — both "provide reasonable efficiency of operation," not "shall not exceed." Designers treat them as targets (the code's own efficiency rationale is real: a 6% drop at 120 V is 7.2 V of lost headroom for the equipment to work with), but an inspection cannot fail a circuit for the note alone. PanelWright's voltage-drop card reports the drop and the smallest size that meets 3% for any run — see the voltage-drop article for the Ch. 9 Table 8 method. EX8 works the pair: a 35 A OCPD on 10 AWG at 100 ft lands at 6.05% (bad), and the core sizes 6 AWG (2.46%) to get under the note.
Every number below is produced by the shipped cores in app.js (pickConductor31016(), nextStdBreaker(), reqBreakerA(), derate31015(), smallConductorCap(), voltageDrop(), sizeForVoltageDrop()) running under node this session — see income-lab/compute_art14.js → calc_21019_cited.json — and is asserted in the public test suite (test/run_tests.js in the public repo). Unless a row says otherwise, conductor picks are in the 75 °C column (110.14(C)(1)(b) terminations; the 60 °C column is the ≤100 A default per 110.14(C)(1)(a) — EX4 shows both) at 30 °C ambient with ≤3 current-carrying conductors.
| Step | Rule | Value |
|---|---|---|
| Continuous load | Article 100 — max current for 3+ h | 16 A |
| Conductor requirement (a) | 210.19(A)(1)(a): 0 + 1.25 × 16 A | 20 A |
| OCPD requirement | 210.20(A): same number → Table 240.6(A) | 20 A |
| Conductor pick | Table 310.16 Cu, 75 °C column: smallest ≥ 20 A | 14 AWG Cu — 20 A |
| 240.4(D) check on the pick | 14 AWG Cu cap = 15 A < 18.75? No — 20 A OCPD needed; 14 AWG capped at 15 A fails | → go to 12 AWG (cap 20 A) |
| Result | 20 A circuit, 12 AWG Cu, 20 A breaker | 16 A = exactly 80% of 20 A |
| Receptacle cross-check | Table 210.21(B)(2): max cord-and-plug load on a 20 A circuit | 16 A |
The three ways of stating the same fact line up: 20 A × 0.80 = 16 A max continuous; 16 A × 1.25 = 20 A required OCPD; Table 210.21(B)(2) says 16 A max on a 20 A circuit. The 2023 code print's own canonical example is exactly this: "Or a 20 amp CB × 80% = 16 amp maximum continuous load." Note the 240.4(D) row: 14 AWG copper has 20 A ampacity in the 75 °C column and would pass the (a) test, but its breaker is capped at 15 A — and 15 A < 18.75 A (the (a) number) would violate 210.20(A). 12 AWG Cu is the smallest size that works; its 20 A cap equals the 20 A requirement.
| Step | Rule | Value |
|---|---|---|
| Load | 960 VA ÷ 120 V (single receptacle — one outlet per circuit, 210.21(B)(1)) | 8 A continuous |
| Conductor requirement (a) | 210.19(A)(1)(a): 0 + 1.25 × 8 A | 10 A |
| OCPD requirement | 210.20(A) → Table 240.6(A): 10 → next standard | 15 A |
| Conductor pick | Table 310.16 Cu, 75 °C: smallest ≥ 10 A | 14 AWG Cu — 20 A |
| 240.4(D) check | 14 AWG Cu cap 15 A ≥ 15 A OCPD | 14 AWG is the smallest usable size |
| Result | 14 AWG Cu, 15 A breaker; receptacle rated ≥ 15 A (210.21(B)(1)) | single-outlet circuit — no (A)(2) floor |
This is the clean case where 14 AWG copper is the right answer: 20 A ampacity clears the 10 A (a) requirement, and the 15 A overcurrent device sits exactly at the 240.4(D) cap for 14 AWG — so the cap is not violated. The contrast with EX1 and EX5 is the whole point of the 240.4(D) trap: 14 AWG works when the required OCPD is 15 A or less (this circuit: 15 A), but the same wire fails the moment the continuous load pushes the required OCPD to 20 A (EX1: 16 A continuous) or 18.75 A (EX5: 15 A continuous) — because the device is then larger than the 15 A cap. And because there is only one receptacle, 210.19(A)(2)'s circuit-rating floor does not apply; 210.21(B)(1) only requires the receptacle rating to be at least the branch-circuit rating (15 A).
| Step | Rule | Value |
|---|---|---|
| Loads | 20 A noncontinuous (lighting) + 15 A continuous (dedicated heater) | — |
| Conductor requirement (a) | 210.19(A)(1)(a): 20 + 1.25 × 15 A | 38.75 A |
| OCPD requirement | 210.20(A) → Table 240.6(A): 38.75 → next standard | 40 A |
| (A)(2) floor | multi-receptacle circuit: ampacity ≥ circuit rating (40 A) | 40 A governs (larger of 38.75 / 40) |
| Conductor pick | Table 310.16 Cu, 75 °C: smallest ≥ 40 A | 8 AWG Cu — 50 A |
| Why not 10 AWG | 10 AWG = 35 A < 40 A (fails (A)(2) floor) AND its 240.4(D) cap is 30 A, so a 40 A OCPD is illegal on it | 10 AWG fails twice → 8 AWG (no (D) cap) |
| Result | 8 AWG Cu, 40 A breaker, 40 A circuit | ampacity 50 A ≥ 40 A floor |
Mixed loads are where the "noncontinuous plus 125% of continuous" split actually does work — only the continuous portion (15 A) is multiplied: 20 + 1.25 × 15 = 38.75 A. That is not a standard OCPD size, so the breaker is 40 A. Two independent reasons 10 AWG (35 A) is out: it does not reach the 40 A (A)(2) circuit-rating floor, and 240.4(D) caps a 10 AWG Cu device at 30 A, so a 40 A breaker cannot legally protect it at all. 8 AWG copper (50 A, no (D) cap) clears both. Note the OCPD (40 A) here is larger than the (a) ampacity number (38.75 A) — 240.6(A) rounds up to the next standard size, and 210.19(A)(2) then pegs the conductor to that circuit rating. The full 240.4 / 240.6(D) mechanics are in the conductor-sizing article.
| Termination column (110.14(C)) | Smallest Cu size ≥ 20 A | Base ampacity used |
|---|---|---|
| 60 °C — equipment ≤100 A / 14–1 AWG (110.14(C)(1)(a)) | 12 AWG Cu | 20 A |
| 75 °C — equipment >100 A or 75 °C-rated terminations (110.14(C)(1)(b)) | 14 AWG Cu | 20 A |
| 90 °C — derating base only, never a termination column (310.15) | 14 AWG Cu | 25 A (base; derating starts here) |
Same load, same 20 A requirement — different wire, depending on the termination column. This is why "14 AWG is 20 A" and "12 AWG is 20 A" are both things you will see quoted: the 60 °C column (the default for equipment rated 100 A or less) needs 12 AWG to reach 20 A, while 75 °C-rated terminations let 14 AWG carry it. The 90 °C column is never a termination column — it exists so derating math (310.15) has a higher base to start from. The shipped core's own notes flag this on every pick: "60 °C column (≤100 A circuits per 110.14(C)(1)(a) unless marked otherwise)" / "90 °C column — use only as a DERATING base (310.15); final ampacity must still respect the termination rating."
| Step | Rule | Value |
|---|---|---|
| Continuous load | Article 100 | 15 A |
| OCPD requirement | 210.20(A): 0 + 1.25 × 15 A → next standard | 18.75 A → 20 A |
| Ampacity pick (a) alone | Table 310.16 Cu, 75 °C: smallest ≥ 18.75 A | 14 AWG Cu — 20 A (passes the (a) test) |
| 240.4(D) cap on 14 AWG Cu | 240.4(D)(4) (2020)/240.4(D) (2023) | 15 A |
| Gate check | cap 15 A < requirement 18.75 A → 210.20(A) violated | 14 AWG NOT usable |
| One size up: 12 AWG Cu | base 25 A; 240.4(D) cap 20 A | cap 20 A ≥ 18.75 A ✓ |
| Result | 12 AWG Cu, 20 A breaker | ampacity 25 A, OCPD 20 A |
This is the single most common branch-circuit sizing mistake, and it exists because the conductor test and the breaker test look at different numbers. 14 AWG copper has 20 A ampacity in the 75 °C column — the (a) test passes. But its overcurrent device is capped at 15 A by 240.4(D), and a 15 A device is less than the 18.75 A that 210.20(A) requires, so the circuit as built would be non-compliant on the OCPD side. One size up, 12 AWG Cu (25 A base, 20 A cap), the cap (20 A) covers the requirement (18.75 A) and the breaker is 20 A. The shipped core emits the cap warning automatically when it picks 14 AWG: "240.4(D): overcurrent device for this size is capped at 15 A" — but note the honest limitation: pickConductor31016() picks by ampacity alone and does not auto-walk the size up on the (D) cap; the OCPD-cap resolution (this EX5) is performed in the article's compute script and asserted in the test suite. The Conductor derating card's check mode does enforce the cap (its "effective ampacity" is min(derated, cap)).
| Step | Rule | Value |
|---|---|---|
| Loads | 21 A noncontinuous + 5 A continuous | — |
| (a) requirement | 210.19(A)(1)(a): 21 + 1.25 × 5 A | 27.25 A |
| Circuit-rating floor | 210.19(A)(2): multi-receptacle → ampacity ≥ rating (30 A) | 30 A governs (larger of 27.25 / 30) |
| Conductor pick | Table 310.16 Cu, 75 °C: smallest ≥ 30 A | 10 AWG Cu — 35 A |
| OCPD | 210.20(A): 27.25 A → next standard (circuit rating sets 30 A) | 30 A |
If you sized by (a) alone, 12 AWG copper (25 A) would seem "close enough" to 27.25 A — it isn't, for two independent reasons: 25 A < 27.25 A fails (a) outright, and the (A)(2) floor demands 30 A anyway. 10 AWG Cu (35 A) clears both. The (A)(2) floor is the reason a "light" multi-receptacle circuit still gets a wire sized to the breaker, not to the load.
| Step | Rule | Value |
|---|---|---|
| Conductor requirement (a) | 210.19(A)(1)(a): 1.25 × 16 A | 20 A |
| (b) check: 14 AWG Cu (base 20 A) | 310.15(B)(1) 36–40 °C @75 = ×0.88; 310.15(C)(1) 8 CCC = ×0.70 | 20 × 0.88 × 0.70 = 12.32 A < 16 A — FAILS |
| (b) check: 12 AWG Cu (base 25 A) | same factors | 25 × 0.88 × 0.70 = 15.4 A < 16 A — FAILS |
| (b) pick: 10 AWG Cu (base 35 A) | same factors | 35 × 0.88 × 0.70 = 21.56 A ≥ 16 A ✓ |
| OCPD | 210.20(A): 16 × 1.25 = 20 A (derating does not change the breaker requirement) | 20 A |
| 240.4(D) check | 10 AWG Cu cap 30 A ≥ 20 A | no cap conflict |
| Result | 10 AWG Cu in the 8-CCC raceway at 40 °C, 20 A breaker | derated 21.56 A |
Watch what didn't change: the 210.20(A) breaker is still 20 A. Derating changes the wire, not the requirement number — the (a) number is computed from the load, and the (b) test just demands the derated ampacity cover the load. The core's derate31015() returns exactly these numbers (baseAmp 35, deratedA 21.56, effectiveA 21.56, passes true for 10 AWG; false for 14 and 12 AWG). This is the v1.16 Conductor derating card's job — enter the load, ambient, and CCC count, and it walks the sizes for you.
| Step | Rule | Value |
|---|---|---|
| Run | 10 AWG Cu (EX6's 30 A circuit), 100 ft one-way, 1∅ 120 V | — |
| Voltage drop | Vd = 2·R·I·D, Ch. 9 Table 8 @75 °C (1.21 Ω/kft) | 7.26 V = 6.05% — over the 5% note |
| Smallest size ≤ 3% note | sizeForVoltageDrop(), target 3% | 6 AWG Cu — 2.95 V = 2.46% |
The ampacity answer (10 AWG, EX6) and the voltage-drop answer (6 AWG, EX8) are different questions with different answers — the same pattern the 215.2 feeder article walks through on the feeder side. Because the 3%/5% figures are informational-note guidance (210.19(A) Info Note No. 3), 10 AWG is code-legal; 6 AWG is the "reasonable efficiency" answer. Which one you install depends on how far the run is and how strictly you apply the note — the tool reports both so the decision is yours. Full treatment in the voltage-drop article.
The panel card sizes every branch circuit against its load and flags continuous loads; the 220.82 card sizes the optional dwelling service and its conductors (Table 310.16, 110.14(C) columns); the 220.61 card sizes the neutral; the Conductor derating card applies the 310.15(B)(1) + 310.15(C)(1) corrections that 210.19(A)(1)(b) demands (the exact EX7 pipeline); and the voltage-drop card checks any run against the 3%/5% informational notes using Ch. 9 Table 8. Breaker picks use the NEC 240.6(A) standard-size list. All results print in the rollup CSV export and the branded PDF project report.
smallConductorCap() (14 Cu→15, 12 Cu→20, 10 Cu→30, 12 Al→15, 10 Al→25) cross-checked against the 2023-NEC-based code print on disk and the conductor-sizing article's 2017/2020/2023 verification. (4) 310.15(B)(1) ambient factors (0.88 for 36–40 °C @75 °C) and 310.15(C)(1) CCC factors (70% for 7–9 CCC) — the shipped core's tables, coordinate-verified from the 2023-NEC print in Session 38. The worked examples EX1–EX8 are computed by the shipped cores under node this session (income-lab/compute_art14.js → calc_21019_cited.json) and asserted in the public test suite — 1023/1023 passing as of v1.16 + article 14 (Session 39). If you find an error in this article or the calculator, the code is plain HTML/JS in the public repo (MIT) — the exact same core that produced every number on this page.