Why Your LED Resistor Drifts: Headroom, Vf Tolerance and a Sagging Battery

You typed in a supply voltage, a forward voltage from a colour preset, and 20 mA. The calculator handed back a resistor value, you soldered it in, and the LED lights. So far so good. Then you build the same circuit again with a different LED from the same bag and it’s noticeably dimmer. Or the battery gets old and the indicator fades weeks before the device stops working. Or the board runs warm and the LED slowly brightens.

None of that is a mistake in the arithmetic. The sizing guide covers the formula, rounding to a standard value, and the power rating, and that math is exactly right — for the numbers you typed. The problem is that two of those numbers are approximations of a real part and a real power source, and how badly the approximation hurts depends on a value the calculator prints but almost nobody reads: the voltage across the resistor.

The row that predicts your error

Look at the results table on the LED Series Resistor Calculator and you’ll see “Voltage across resistor” sitting above the resistance. That’s the supply left over after the LEDs have taken their share — the headroom. It has no units anyone shops for and you never solder it in, but it’s the single number that tells you how much your circuit will drift.

Here’s the relationship, and it’s exact rather than a rule of thumb. If the real forward voltage differs from the one you typed by some amount, the fractional error in your current is:

current error = (number of LEDs × Vf error) ÷ headroom

The resistor doesn’t care what the LEDs are doing. It only sees the leftover voltage, and it converts that leftover into current linearly. So any voltage that goes missing from the headroom comes straight out of your current — and the smaller the headroom, the larger a share that missing voltage represents.

The same LED, two supplies, wildly different stability

Take a white LED, nominal Vf 3.2 V, target 20 mA, and drive it two ways.

From 5 V. Headroom is 1.80 V. The calculator asks for 90 Ω, rounds up to 100 Ω, and predicts 18.0 mA. Now suppose the actual part measures 3.4 V instead of 3.2 V — well within the spread you’d find on a datasheet, sometimes within a single reel. Headroom drops to 1.60 V and the current falls to 16.0 mA, an 11% miss.

From 12 V. Headroom is 8.80 V, the resistor is 470 Ω, predicted current 18.7 mA. Same 0.2 V error on the same part: current moves to 18.3 mA. That’s a 2.3% miss — invisible.

Identical LED, identical target, identical error in the input. One circuit is five times more sensitive than the other, and the only thing that changed is how much of the supply the resistor got to keep. This is why two boards built from the same bag of parts can be visibly different brightnesses on a 5 V rail and perfectly matched on 12 V.

A useful habit: after the calculator gives you a value, glance at the headroom. If it’s under about a volt, or under roughly a fifth of your supply, treat the predicted current as a rough centre of a range rather than a number you can count on.

A battery is not a supply voltage

The supply box takes one number, but a cell doesn’t hold one number. A fresh alkaline 9 V measures closer to 9.6 V; by the end of its useful life it’s down around 7 V and still lighting things. Take the 390 Ω the sizing guide landed on for a red LED and walk that battery down:

Battery state	Supply	Current
Fresh	9.6 V	19.5 mA
Nominal	9.0 V	18.0 mA
Nearly flat	7.0 V	12.8 mA

That’s a 34% swing in current over the life of one battery, with the resistor doing exactly what it was told. The LED gets dimmer as the cell drains, which is often fine for an indicator and not fine at all for anything you’re using to illuminate or signal. The fix isn’t a different resistor — it’s accepting the range, or moving to a regulated supply, or driving the LED harder than you’d like when fresh so it’s still adequate when flat. Whichever you choose, run the calculator at both ends of the battery’s range rather than at its nominal label, and check how long that current actually lasts with the Battery Runtime Calculator.

The same applies to rails you think are fixed. A car’s “12 V” system sits near 13.8 V with the engine running. That 470 Ω resistor from the example above now passes 22.6 mA — above the 20 mA you designed for — and its dissipation climbs from 165 mW to 239 mW, because power rises with the square of the headroom.

What headroom costs: everything it doesn’t give the LED

Stability isn’t free. The LED converts its share of the voltage into light; the resistor converts its share into heat. Nothing else happens. So the fraction of the supply the LED takes is the efficiency of the circuit:

• White LED on 12 V: the LED gets 27% of the power. 73% is heat.
• Red LED on 9 V: the LED gets 22%.
• White LED on 5 V: the LED gets 64%.

That inverts the advice above. The 12 V circuit that was five times more stable is also burning nearly three quarters of its energy in a resistor — a terrible trade on anything battery-powered, and a real thermal problem if you have dozens of indicators. The 5 V circuit that drifts is the one that’s actually using its battery well.

You can see the consequence in the recommended rating the calculator prints. The 5 V version dissipates 32 mW and asks for a 1/8 W resistor. The 12 V version dissipates 165 mW and asks for a 1/2 W part — four times the wattage, physically larger and more expensive, to light the same LED to the same brightness.

Stacking LEDs in series buys efficiency and sells stability

The “LEDs in series” box is the lever that fixes the efficiency problem, and it’s more useful than it looks. Three white LEDs on 12 V take 9.6 V between them, leaving 2.4 V of headroom. The calculator lands on exactly 120 Ω, a full 20.0 mA, and the LEDs now receive 80% of the power — three times the light of the single-LED version from the same current draw, with the resistor barely warm at 48 mW.

But re-read the error relationship. The Vf error is multiplied by the number of LEDs and divided by a headroom you just shrank. Three LEDs each 0.2 V above nominal cost you 0.6 V out of 2.4 V — a 25% drop in current, from 20.0 mA to 15.0 mA. The stack is four times more efficient than the single LED on 12 V and eleven times more sensitive to part variation.

This is the real design tension, and it has no clever resistor solution. Long series strings with tight headroom are efficient and twitchy; short strings with fat headroom are stable and wasteful. If you need both — matched brightness and efficiency — you’ve outgrown a resistor and want a constant-current driver, which senses the actual current and adjusts, making Vf variation irrelevant by design.

The temperature term you never type in

There’s a third input drift the calculator can’t model because it doesn’t exist until the circuit is running. Forward voltage falls as the junction heats, roughly a couple of millivolts per degree C. A LED that warms 40 °C above ambient sheds something like 0.08 V, which quietly raises the headroom and therefore the current, which produces more heat.

With generous headroom this is self-limiting and irrelevant — 0.08 V out of 8.8 V is under 1%. With 2.4 V of headroom across a three-LED stack it’s a few percent, still fine. The reason it’s worth knowing is that it always pushes in the same direction as every other low-headroom problem: the tighter you run, the more the circuit’s behaviour depends on things you didn’t enter into any box.

Why there’s no box for parallel LEDs

The calculator counts LEDs in series and offers no way to say “four LEDs in parallel behind one resistor.” That omission is deliberate, and it’s the same story as everything above pushed to its limit.

Put two LEDs in parallel and they’re forced to the same voltage — but they don’t have the same Vf. The one with the lower forward voltage conducts more, heats up, drops its Vf further, and takes an even larger share. The current doesn’t split evenly; it concentrates, and it concentrates on the part least able to take it. One LED runs bright and hot while the other sits dim, and the bright one is the one that fails.

The correct build is one resistor per parallel branch, and that’s exactly what the tool supports: run it once per branch, entering the number of LEDs in that branch. Four separate branches of one LED each means four resistors, and each branch gets its current set independently, immune to how its neighbours’ forward voltages happen to have landed. The extra resistors cost pennies; the shared resistor costs an LED.

A pre-solder checklist

Before you commit the value: read the headroom, not just the resistance. If it’s under a volt or under a fifth of the supply, expect double-digit percentage swings in current between parts, and don’t rely on brightness matching. Run the calculator at both ends of your supply’s real range, not its label. Multiply your Vf uncertainty by the number of LEDs in the string before deciding whether the headroom is enough. And if you’re wiring more than one LED, decide series or separate branches — never a shared resistor across parallel parts.

Then open the LED Series Resistor Calculator, enter your numbers, and read the whole results table: the exact resistance, the nearest standard value, the real current you’ll get at it, and — first and most useful — how many volts the resistor is holding back.