IPC-2221 vs IPC-2152: Where Your Trace Width Number Actually Comes From

If you have ever typed a current and a temperature rise into a trace width calculator and copied the answer onto a board, this is about where that answer came from. It came from a test run in 1956, on a Navy contract, whose own authors labelled the results "Tentative" and asked for funding to do it properly.

That funding never arrived. The charts got redrawn, republished, and eventually the word "Tentative" fell off. They are still the numbers behind a great many quick trace width tools, including ours. Below is what they actually measured, the one part of them that was never measured at all, and what changed when the work finally got done properly in 2009.

The paper trail back to 1956

Michael Jouppi chaired the IPC task group that eventually replaced these charts, and in a paper published by IPC he sets out the origin plainly: "In 1955 the United States National Bureau of Standards was funded by the Department of the United States Navy to develop a method for evaluating conductor current carrying capacity in printed circuits. It was at a time when printed circuits were first being introduced as a technology."

His reference for the underlying data is NBS Report #4283, "Characterization of Metal-insulator Laminates", dated May 1, 1956, commissioned by the Navy Bureau of Ships. On the result: "The end result was a chart labeled 'Tentative'. Although it was published as a chart for sizing external conductors, it eventually became Mil-Std-275."

Douglas Brooks and Johannes Adam trace the same lineage in Trace Currents and Temperatures Revisited (April 2015), and add the detail that stings: "The empirical data were not very well controlled (because of limited resources), and the resulting charts, when published, were labeled 'Tentative.' The authors recommended that funding be provided for a more detailed, more carefully controlled study, but such funding was never forthcoming."

Then: "Through the years the charts were redrawn and republished, and somewhere along the line the word 'Tentative' was dropped." Their reading of the publication chain is MIL-STD-1495 in 1973, MIL-STD-275E in 1984, then IPC-D-275, then IPC-2221.

None of that makes the charts worthless. Brooks is fair about why they survived: "In hindsight, the best thing they had going for them was the test of time. They apparently were appropriately conservative because few board failures were traced back to using them."

The internal-trace chart was never measured

This is the part worth knowing before you use any calculator's internal setting.

Brooks and Adam: "Originally, there were two sets of charts, one for external traces and one for internal traces. The empirical data only applied to the external traces. The internal trace charts were derived by de-rating the external charts by a factor of two, on the expectation that the internal traces would not cool as well as the external traces would, and would therefore be hotter."

So the internal chart is not a measurement. It is the external measurement, halved, on a guess about buried copper being hotter.

That halving is still sitting in the arithmetic today, and you can see it in our own source. The IPC-2221 current equation is I = k × ΔT^0.44 × A^0.725, and the calculator carries two constants: k = 0.048 for external traces and k = 0.024 for internal ones. Exactly half, exactly as described.

What that does to a width is bigger than most people expect, because area enters the equation raised to 0.725. Halving the allowed current means the area has to grow by 2^1/0.725, or a factor of 2.60.

Run 2 A at a 10°C rise on 1 oz copper through our calculator and the external answer is 0.78 mm (30.8 mil). Flip it to internal and it becomes 2.03 mm (80.0 mil). Same current, same temperature target, a trace two and a half times wider, on the strength of an assumption nobody tested.

What IPC-2152 found when the test was finally done

IPC-2152, "Standard for Determining Current Carrying Capacity in Printed Board Design", was published on 16 September 2009, 97 pages, covering "how thermal conductivity, vias, copper planes, power dissipation and printed board material and thickness all factor into the relationship between current, conductor size, and temperature."

Brooks calls it "the best researched, best controlled, most thorough study ever made of trace currents and temperatures." And the headline finding is that the 1956 guess about internal traces was backwards.

In his words: "It turns out the internal traces cool almost as well or better than do the external traces. The IPC external 2 oz curves show higher temperatures than do the internal curves for the same traces. That is because it turns out the board materials conduct heat away from the trace better than the air does. This is the one assumption the original researchers got very wrong."

Adam reached the same conclusion independently through 3D simulation, writing in Bodo's Power Systems (October 2011) that simulations "disprove the myth of hot internal traces: an internal trace can be even cooler than an external trace, because of better heat spreading above and below."

Jouppi goes further and explains what the old internal chart was really showing. Looking at the original test data, he notes that some conductors were stripped out of the boards entirely and tested bare, and that this is the line the internal chart tracks: "The IPC-2221 internal conductor sizing chart, Figure 3, actually represents a conductor in free air."

A bare wire hanging in air, in other words, labelled for fifty years as the guidance for copper buried inside a laminate. That is why, as Jouppi puts it, IPC-2152 "answers questions that are asked on every Continent about why the internal chart that has been used since 1955 does not compare with engineering calculations and test data."

The part where the sources disagree, and why

Read around this topic and you will find confident claims in both directions. Some say IPC-2152 lets you use narrower traces than IPC-2221. Others say it is the more conservative of the two. Both are in the sources above, and both are correct about different things.

Jouppi writes that "The old charts oversize conductors." Adam writes that "the new IPC-2152 values are close to 'DN' correlations and is therefore more conservative than the old 2221." Those look like a flat contradiction until you check which chart each is talking about.

For internal traces, IPC-2221 massively oversizes, for the reason above. For external traces it goes the other way. Brooks quantifies it: "The IPC 2152 data result in currents approximately 25% lower than those shown in the original IPC 2221 set of curves."

The reason is what each standard assumed was underneath the trace. IPC-2152's baseline is deliberately bare: Jouppi describes it as "a 0.07 inch thick polyimide test vehicle", with "no copper planes, suspended in still air as well as in vacuum." Nothing to conduct heat away.

Adam's simulation shows how much that single choice moves the answer. Modelling a 1.4 mm wide, 35 µm trace carrying 5 A, he got a 26 K rise with copper cladding on the opposite face of the board, and 53 K with that cladding removed. The first figure matches the IPC-2221 external chart. The second is close to IPC-2152 and the older Design News data.

His coefficient table makes the same point in one line. The multiplier on predicted temperature rise is 3.6 for a bare IPC-2152 or Design News board, 1.7 for an IPC-2221 board, and 1.0 for a board with two internal 35 µm copper planes. Same trace, same current, a spread of more than three to one depending only on what copper sits nearby.

So the two standards are not really competing estimates of one quantity. IPC-2221's external chart is a measurement of a trace on a clad board. IPC-2152's baseline is a measurement of a trace with nothing helping it, plus the correction factors to add the help back in.

What actually moves the temperature

Once you have seen the spread, the useful question stops being which standard to quote and starts being which features of your board are doing the cooling. The figures below are all from the sources above.

• Copper planes. Jouppi's worked example: "a conductor estimated to have a 30C rise calculated using the IPC-2152 charts will have a temperature rise closer to 9C when a 2-oz copper plane (5-in x 5-in [25sq.in]) is present in the design located 0.005-inch away from the trace."
• Neighbouring traces. IPC-2152 treats conductors within one inch of each other as parallel. Two traces each sized on their own for a 10°C rise, then placed 0.10 inch apart, came out at "a little more than 17C rather than the desired 10C" in Jouppi's example. Close them to a few thousandths of an inch and it is 18C.
• Trace length. Brooks and Adam modelled a 1 oz, 200 mil wide external trace at 15 A. At 6 inches long the rise was 94.7°C. At 2 inches, 81.5°C. At 1 inch, 64.6°C. Short traces dump heat out through the pads at each end.
• Copper weight, counterintuitively. Jouppi notes "for the same size conductor in terms of cross-sectional area, a conductor of 1-oz copper runs cooler than heavier weight coppers for the same applied current." Thicker copper at equal area means a narrower trace, and width is what does the cooling.
• Board thickness and material. Between polyimide and FR4 the difference "had little impact on conductor temperature rise", but thinning the board increases it.

That last group is why Brooks and Adam end up where they do, concluding that "trace current/temperature relationships are too complex to represent with equations or graphs; thermal simulation models are required."

So how should you use a quick calculator

Their conclusion is right and also unhelpful if you are laying out a small board this evening and are not about to buy a thermal simulator. Here is the honest middle.

Our PCB Trace Width Calculator implements IPC-2221, and it says so on the tin. That is the correct choice for a fast sanity check, and Altium's own write-up of the same formula reaches the same verdict, describing IPC-2221 results for a modern board with copper planes and pours as "very conservative" and "probably an overestimate" of the width you need.

A few practical adjustments follow from everything above.

Use the external setting as your working number even for inner-layer traces. Since internal traces measure as cool as or cooler than external ones, the external figure is not the reckless choice its label suggests. The internal setting is not unsafe, it is just wasteful, and on a dense board that 2.6× width penalty costs you routing space you did not need to spend.

Do not mistake that external number for a worst case, though. It is the one part of IPC-2221 that IPC-2152 found optimistic rather than conservative, by Brooks' 25% on current. The fix is to ask the calculator for more current than you actually need: divide your real current by 0.75 and enter that. For the 2 A example, enter 2.67 A, and the answer moves from 0.78 mm to 1.16 mm.

Those two adjustments belong together. External setting alone is optimistic on a bare board, and internal setting alone is wasteful on any board. External plus the derating lands close to what IPC-2152 would actually tell you.

Treat the result as a floor for a single isolated trace, then look at what is around it. Two power traces side by side are not two independent problems, they are one hotter problem. A trace that gets shorter or picks up a nearby plane gets cooler than any chart will tell you.

And keep the temperature rise you asked for in view. A 10°C rise on a bench at 25°C is a very different thing from a 10°C rise inside a sealed enclosure that already sits at 60°C. The charts only ever gave you the rise, never the final temperature.

If you want the real numbers for an inner layer, a vacuum environment, or a board with planes at a known distance, IPC-2152 is the document with them, and there is no free equation that stands in for its charts. For everything else, a conservative old number you understand the provenance of beats a modern one you do not.