AWG wire gauge and current: what actually sets the limit

A wire does not have a current rating. Its insulation does.

Copper melts at 1083 °C and is a perfectly good conductor at 400 °C. Nothing in a stranded copper conductor cares about 10 A. What fails first is the plastic around it, and the plating on the strands underneath — so an ampacity figure is never a property of the wire alone. It is the answer to a question with four more terms in it: how hot the surroundings already are, how much hotter the insulation may legally get, how many other current-carrying wires are pressed against it, and whether there is any air to carry the heat away.

That is why the same 22 AWG wire is rated 5 A in one place and 2.5 A in another, and why neither number is wrong. It also means the number you actually need is usually not on the chart at all — but the thing that decides most low-voltage wiring is not heating anyway. It is voltage drop, and that one is exact.

The gauge number is a definition, not a table

American Wire Gauge looks like a lookup table and is not. It is two fixed points and a geometric series between them: #36 is 0.005 inches, #0000 is 0.460 inches, and there are exactly 39 steps from one to the other. Every gauge follows:

$$d(n) = 0.005 \cdot 92^{\frac{36 - n}{39}} \text{ inches}$$

The ratio between adjacent gauges is therefore $92^{1/39} = 1.1229$, and the whole standard is that constant. Nothing needs to be memorised and no table needs to be transcribed — which matters, because a transcribed table is a place for a typo to live.

The gauge number runs backwards because it counts drawing operations: a wire was pulled through successively smaller dies, and the number is how many times. More pulls, thinner wire, bigger number. Gauges wider than #1 ran out of positive numbers and went to 0, 00, 000, 0000 — numbered 0, −1, −2, −3 when you compute with them.

Two ratios worth carrying in your head

The geometric definition gives two rules that are exact enough to use without a calculator, because area goes as diameter squared:

• Three gauges is a factor of two in copper. $1.1229^{6} = 2.005$.
• Ten gauges is a factor of ten. $1.1229^{20} = 10.16$.

So 14 AWG has twice the copper of 17 AWG, and ten times the copper of 24 AWG. Both are accurate to better than 2 %, which is far tighter than anything else in this problem.

Resistance is exact. Ampacity is not.

Once the geometry is fixed, resistance follows from copper’s resistivity with nothing left to argue about:

$$\frac{R}{L} = \frac{\rho_{Cu}}{A}\left(1 + \alpha_{Cu}\,(T - 20)\right)$$

with $\rho_{Cu} = 1.68 \times 10^{-8}\ \Omega\text{m}$ and $\alpha_{Cu} = 0.00393\ \mathrm{K^{-1}}$ — the same constants the trace width calculator uses, so a wire and a trace on this site never disagree about what copper does.

That temperature term is not a rounding error. A conductor running 100 K above where it was characterised carries 39 % more resistance, so it drops 39 % more volts and dissipates 39 % more power at the same current. Wire that is already hot gets hotter faster than a linear intuition suggests.

What an ampacity rating actually measures

NASA’s Engineering and Safety Center spent a whole assessment on this question. Its re-architecting of the NASA wire derating approach opens the discussion by noting that “there is no general NASA standard for wire current rating/derating, or how to define the temperature of a wire”, and quotes the JSC memorandum for Space Shuttle payloads (NASA/TM-102179) as the example of a rating written in terms of the insulation. The allowable current in a selected environment is:

the amount of current required to raise the insulation temperature from that of the wire in a nonconducting state (insulation temperature is equal to ambient) to the maximum rated temperature of the insulation.

The assessment itself then takes the other basis — it assumes the published rating curves (AS50881, JPL D-8208) are tied to the temperature of the conductor, not the insulation, which is the temperature its thermal models solve for and its resistance-based measurements read. Conductor and jacket are in contact, so the two temperatures move together, and the accounting is identical either way.

Read the definition as a budget rather than a rating. The material has a ceiling. The ambient has already spent part of the distance to it. Whatever is left is the temperature rise your current is allowed to cause — and that remainder, not the wire, is what the ampacity number describes.

The consequence is the part people skip. A wire rated for 5 A at 25 °C ambient has no such rating inside an enclosure at 85 °C. The copper has not changed; the budget has been spent before any current flows. This is the same accounting as θJA and the board it is measured on: a thermal number quoted without its environment is not a number.

The ceiling belongs to the insulation and the plating

The same NESC report is explicit that two materials set the limit, and neither is the conductor:

The wire conductor and insulation materials both limit the actual current rating of the wire system. An aerospace grade copper conductor is plated with tin (rated at 150 °C), silver (rated at 200 °C), or nickel (rated at 260 °C) to provide a stable conductive service over the range of expected temperatures.

Insulation ratings sit in the same band — the report gives XL-ETFE at 200 °C and TKT at 260 °C. So a wire’s ceiling is whichever of its plating and its jacket gives up first, and a silver-plated conductor inside a 150 °C jacket is a 150 °C wire.

This is also why “tinned copper” on a bill of materials is a thermal specification and not just a solderability one.

The same wire has at least five ratings

Bundling is the term that moves the number most, and it moves it further than most engineers expect. Alpha Wire’s wire ampacity overview gives a chart current per gauge and temperature rise, then a multiplier for how many current-carrying conductors are bundled together:

conductors	multiplier
1	1.6
2 to 3	1.0
4 to 5	0.8
6 to 15	0.7
16 to 30	0.5

A single conductor in free air gets 1.6× the chart value. The same wire in a 30-way loom gets 0.5×. That is a spread of 3.2 to 1 on identical copper, decided entirely by what is next to it. A wire that was comfortable on the bench, run singly, can be over its rating the moment the harness is laced.

The physical reason is that the interior of a bundle has nowhere to send its heat except through the wires around it, each of which is also generating heat. The NESC work models this as a contact conductance between elements, and found the parameter varied enough between test campaigns to “raise serious questions about the characterization and repeatability” of it. If NASA cannot pin the bundle term down repeatably, a chart multiplier is doing well to be within a factor of two.

Why these charts are measured rather than derived

It is tempting to derive ampacity from first principles. Balance the heat generated against the heat lost: $I^2R$ into a surface that sheds heat over a perimeter proportional to $d$, with $R \propto 1/d^2$. That gives

$$I \propto \sqrt{\Delta T \cdot d^{3}}$$

The scaling is sound and the absolute numbers are not. Normalising that law to Alpha Wire’s own 14 AWG point (8 A at a 10 K rise) and asking it to predict their 22 AWG point (35 K rise) gives 3.7 A where the published chart allows 5 A — 26 % low.

The shape is right; the number is not. Real wires have insulation with its own conductivity and emissivity, strands with air between them, and a convection coefficient that changes with temperature and orientation. That is why the standards NESC examined — AS50881, JPL D-8208 — are empirical, and why the report describes them as adding “considerable design margin”. Use the charts. Do not re-derive them and believe the answer.

Voltage drop usually binds first

For most low-voltage wiring, none of the above is the constraint. Heating is a distant second to the volts you lose on the way.

$$V_{drop} = I \cdot \frac{R}{L} \cdot 2L_{\text{run}}$$

The factor of two is the mistake worth naming: current goes out and comes back, so a 5 m run is 10 m of copper. Half of all drop miscalculations are that missing 2.

At 1 A a 250 mV budget is spent after 6.1 m on 18 AWG, 2.4 m on 22 AWG, and 0.96 m on 26 AWG. The wire gauge calculator works the drop and the bundle derating together, and reports the gauge that would bring the drop inside 5 % of the rail. None of those wires is remotely warm at 1 A. The gauge was decided by the regulator’s tolerance, not by temperature — and if the load is a sensor whose reading depends on its supply rail, the drop is an accuracy error before it is anything else.

A wire rating never transfers to a PCB trace

Wire ampacity charts and PCB trace charts answer the same question about different geometry, and the numbers do not carry across. A round conductor sheds heat in every direction into moving air. A trace is a thin ribbon bonded to laminate, with copper planes nearby that act as heatsinks and neighbouring traces that do not.

Matching 24 AWG — 0.205 mm² of copper — in 1 oz foil takes a trace 5.85 mm wide. Anyone sizing board copper by remembering a wire chart will be out by a large factor in whichever direction is least convenient. The trace width calculator applies IPC’s curves, which were measured on laminate.

Continuous current and fault current are different questions

Everything above is steady state: the wire has reached a temperature and stays there. A short circuit is the opposite regime. The event is over in milliseconds, no meaningful heat leaves the copper in that time, and the question is not how hot the jacket gets but whether the conductor survives at all. The calculator treats that no-heat-leaves assumption as fair out to about 5 s; beyond that the jacket and whatever the wire touches start to carry heat away, and the fusing figure turns conservative rather than wrong.

That is the adiabatic problem Onderdonk’s equation solves, and the fusing current calculator works it for board copper. Sizing a conductor for a fault using an ampacity chart, or sizing it for continuous duty using a fusing equation, both give answers that are wrong by more than an order of magnitude.

A worked example

A 22 AWG sensor cable, 2 A, 5 m each way, in a loom of eight conductors, inside a 60 °C enclosure. The gauge is chosen so the heating check can use a published chart point rather than an interpolated one.

Geometry     d = 0.005 × 92^((36−22)/39) = 0.0253 in = 0.644 mm
             A = π(0.644/2)² = 0.326 mm²

Resistance   R/L = 1.68e−8 / 0.326e−6 = 0.0516 Ω/m at 20 °C

Drop         V = 2 A × 0.0516 Ω/m × 10 m = 1.03 V

Heating      published chart point: 5 A at a 35 K rise
             bundle of 8 → × 0.7  →  3.5 A allowed, and 2 A is drawn
             60 °C ambient + 35 K rise = 95 °C, inside a 105 °C jacket

Budget       105 °C jacket − 60 °C ambient = 45 K left before any current flows
             the chart's 35 K rise takes 35/45 = 78 % of it; 10 K remains

The thermal answer is that 2 A is comfortable — but look at where the budget went. The 60 °C enclosure spent 57 % of the 105 °C jacket rating before any current flowed, and the chart’s 35 K rise takes 78 % of the 45 K that were left. The drop answer is that 1.03 V has been thrown away — a fifth of a 5 V rail, and a third of a 3.3 V one. The wire passes the test everyone runs and fails the one that matters.

Going to 20 AWG is 1.59× the copper and brings the drop to 0.65 V; 18 AWG is 2.53× and brings it to 0.41 V; 16 AWG reaches 0.26 V. None of those changes was made for temperature, and the thermal check passed before any of them.

The order to size a wire in

1. Compute the drop first, both conductors, at the highest current the load actually draws. On anything below 24 V this decides the gauge most of the time, and the calculator does it from the gauge, the current and the run.
2. Then check heating — chart current for the gauge and the rise you can afford, times the bundle multiplier.
3. Subtract the ambient from the jacket rating before believing any rise. The budget is what is left, not what the chart assumed.
4. Check the plating and the jacket separately. The lower of the two is the wire’s ceiling.
5. Size fault protection separately. It is an adiabatic problem, not a steady-state one.

The first step is exact arithmetic. The second is a measured chart with a factor-of-two bundle term in it. Spending effort on the second while skipping the first is the common failure.

Sources

• NASA/NESC, Re-Architecting the NASA Wire Derating Approach, NESC-RP-17-01264 — the JSC ampacity definition it quotes (and the conductor-temperature basis it adopts instead), the plating and insulation temperature ratings, the bundle conductance modelling, and the finding that the established standards are empirical and carry considerable margin.
• NASA-STD-8739.4A, Crimping, Interconnecting Cables, Harnesses, and Wiring — the workmanship standard the derating work sits alongside.
• Alpha Wire, Wire Ampacity Overview — the conductor-count multipliers and the two published chart points used above.

Every diameter, area, resistance and drop on this page is computed from the AWG definition and copper’s resistivity in scripts/figures/wire-gauge-current-rating.figures.mjs, and checked against published gauge dimensions in tests/calc/wire-gauge.test.ts. No gauge table was transcribed.