Shunt Resistor Calculator
Shunt Resistor Calculator
Full-scale current and a target sense voltage give the shunt — and then everything that goes wrong: the nearest standard value and its error, the dissipation and the wattage you really need, the temperature the shunt reaches, the temperature-coefficient error at that temperature, and how much a few milliohms of lead and solder cost you without a four-terminal connection.
Shunt value, dissipation, self-heating and the errors
10 A full scale into 50 mV, on a 1 W 2512 rated at 70 °C and derating to zero at 170 °C, with 110 ppm/°C and 0.5 mΩ of lead in the sense loop
Ohm’s law, and then the thermal loop it sits in
Rth = (Tzero power − Trated) ÷ Prated
Rhot = R·(1 + α(Tambient − 25)) ÷ (1 − α·R·I²·Rth)
Tshunt = Tambient + I²·Rhot·Rth
gain = VADC full scale ÷ (I · Rhot)
two-terminal error = Rleads ÷ Rshunt
- alpha
- the temperature coefficient in parts per million per degree. Manganin is about 10, a good metal-strip chip resistor 35 to 110, and a very low-value one 400
- Rth
- thermal resistance. A linear derating curve is exactly this: the temperature span divided by the rated power
- the denominator
- α·R·I²·R_th is the loop gain of the self-heating. At 1 the resistance is unbounded — thermal runaway
- R leads
- everything in the voltage-sense path that also carries load current. A four-terminal connection removes it by construction
Worked example
10 A full scale into 50 mV, on a 1 W 2512 rated at 70 °C and derating to zero at 170 °C, with 110 ppm/°C and 0.5 mΩ of lead in the sense loop
50 mV ÷ 10 A = 5 mΩ; the nearest E24 value is 5.1 mΩ, which is 2.0% high — a fixed scale error you calibrate out once, or avoid by going to E96
At full scale it dissipates 10² × 5.1 mΩ = 510 mW, which is 51% of a 1 W part — so at a 50% derating you want a part rated 1.026 W or more
The derating curve gives the thermal resistance directly: (170 − 70) ÷ 1 W = 100 °C/W, so the shunt settles at 76.3 °C — 51 °C above ambient, on a part whose ambient you probably thought was the specification
At 76.3 °C its resistance has risen 0.564%, so the sense voltage at full scale is 51.29 mV rather than 51 mV — and the amplifier needs a gain of 64.3 to reach 3.3 V
Finally the layout: 0.5 mΩ of track and solder inside the sense loop is 9.8% of a 5.1 mΩ shunt, which dwarfs everything above. A four-terminal connection removes it by construction; nothing else does
Temperature coefficients of real current-sense resistors
| Part | Resistance | TCR | Notes |
|---|---|---|---|
| Vishay WSL2512 | 0.5–0.99 mΩ | ±400 ppm/°C | 1 W at 70 °C, derating to zero at 170 °C |
| Vishay WSL2512 | 1–2.9 mΩ | ±275 ppm/°C | same package |
| Vishay WSL2512 | 5–6.9 mΩ | ±110 ppm/°C | the default used here |
| Vishay WSL2512 | 50–100 mΩ | ±75 ppm/°C | same package |
| Vishay WSK2512 | 5 mΩ–0.2 Ω | ±35 ppm/°C | the improved-stability version, four-terminal, 1 % down to 0.5 mΩ |
| Isabellenhütte MANGANIN alloy | — | ±10 ppm/K | between +20 and +50 °C; 43 µΩ·cm resistivity and −1.0 µV/K against copper, which is why precision shunts are made of it |
High side or low side
| High side | Low side | |
|---|---|---|
| Where | between the supply and the load | between the load and ground |
| Common-mode voltage the amplifier sees | the full bus voltage, plus transients | near ground |
| Load’s ground | untouched | lifted by the sense voltage, and it moves with the current |
| A short from the load to chassis | shows up as current | invisible |
| Amplifier needed | a dedicated current-sense amplifier rated for the bus | almost any op-amp |
| Typical use | battery packs, motor drives, anything where a ground fault matters | low-voltage supplies, where the ground shift is tolerable |
The easy calculation and the four hard ones after it
A shunt is a resistor you have deliberately put in the path of the current you want to measure, chosen so the voltage across it is something you can amplify. The value is Ohm’s law and takes a second. Everything that makes current sensing difficult happens afterwards.
Dissipation goes as the square of the current. Halving the sense voltage halves the power, which is why the industry has moved from 100 mV panel-meter shunts to 10 or 25 mV parts with a dedicated amplifier — and why the traditional answer to a big current is a big piece of metal with its own bolt terminals rather than a chip resistor. The number to watch is not the wattage but the temperature: a 1 W 2512 derating linearly from 70 °C to 170 °C is telling you it has a thermal resistance of 100 °C/W, so half a watt puts it 50 °C above the board it is soldered to. That is a component running at 75 °C in a 25 °C room, and the specification sheet never says so in those words.
Then the resistance changes. Every shunt has a temperature coefficient, and at the temperature its own dissipation takes it to, that coefficient is usually the largest error in the measurement — larger than the tolerance, larger than the amplifier’s offset, and, unlike both of those, it moves with the load so a one-point calibration cannot remove it. Worse, the loop is regenerative: hotter means more resistance means more power means hotter. It converges for any sane part, and the closed form on this page makes the convergence condition explicit — its denominator is 1 − α·R·I²·Rth, and at zero you have thermal runaway. This is why precision shunts are manganin, at about ±10 ppm/K, rather than the ±110 to ±400 ppm/°C of an ordinary low-value chip resistor.
And then the layout undoes it all. A 5 mΩ shunt with half a milliohm of track, via and solder inside the voltage-sense path reads 10% high, and that half milliohm is copper, which drifts at 3,900 ppm/°C — ten to four hundred times worse than the shunt it is corrupting. The fix is a four-terminal, or Kelvin, connection: two terminals carry the load current and two entirely separate ones carry no current at all and sense the voltage at the resistive element itself. Four-terminal shunts have the pads for it; a two-terminal part can be Kelvin-connected by taking the sense tracks from the very inner edges of its terminations and routing them as a tight differential pair that joins nothing else on the way. At 5 mΩ this is not a refinement, it is the difference between an instrument and a guess.
High side or low side is a separate decision with no right answer. A low-side shunt is trivial to amplify and puts a resistor in the ground return, so the load’s ground now sits above system ground by the sense voltage and moves with the current — and a fault current that escapes to chassis never passes through the shunt at all. A high-side shunt leaves the ground alone and sees every fault, at the cost of an amplifier that has to work with its inputs sitting at the bus voltage and survive whatever the switching node does on top of it.
For a transformer instead of a resistor, which gives isolation and no dissipation but works only on AC, see the CT burden resistor calculator. For the amplifier see the non-inverting amplifier calculator, for the digitiser the ADC resolution calculator, for the thermal side the heatsink thermal resistance calculator, and for the same problem in a bridge, the load cell calculator.
Frequently asked questions
How do I calculate a shunt resistor value?
Divide the sense voltage you want at full scale by the full-scale current. 50 mV at 10 A is 5 mΩ. The hard part is choosing the sense voltage: more gives a better signal against the amplifier’s offset and noise, and costs power as the square of the current.
What power rating does a shunt resistor need?
At least I² × R, and in practice about twice that, because a resistor at its full rating is at the temperature where that rating derates to zero. Work with the temperature rather than the wattage: the derating curve gives you a thermal resistance, and that tells you how hot the part will actually get.
Why does my current reading drift as the load warms up?
Because the shunt is warming up too, and its resistance is rising with it. At 110 ppm/°C and a 50 °C rise that is 0.55%, which is more than the tolerance of the part. It is load-dependent, so it cannot be calibrated out at one point — a lower-TCR part, a lower sense voltage or more copper under the shunt are the real fixes.
What is a four-terminal (Kelvin) connection and do I need one?
Two terminals carry the load current and two separate ones sense the voltage at the element, carrying no current, so no lead or solder resistance appears in the measurement. You need one whenever the resistance in the sense path is a significant fraction of the shunt — which for anything below about 10 mΩ means always. Half a milliohm of track on a 5 mΩ shunt is a 10% error.
Should I sense on the high side or the low side?
Low side if the load will tolerate its ground moving and you do not care about faults to chassis; high side otherwise. High side needs an amplifier whose common-mode range covers the bus voltage and its transients, which is the whole reason dedicated current-sense amplifiers exist.
Why are precision shunts made of manganin?
Two reasons, both about temperature. Its resistance barely changes with it — Isabellenhütte specify ±10 ppm/K for MANGANIN between 20 and 50 °C against hundreds of ppm for ordinary resistance alloys — and its thermal EMF against copper is about −1 µV/K, so the junctions at each end do not generate a spurious signal when one is warmer than the other. Against a sense voltage of tens of millivolts, microvolts matter.
Related calculators
References
- Vishay Dale, Power Metal Strip Resistors, WSL. 1 W at 70 °C for the 2512, derating linearly to zero at 170 °C, and the TCR table reproduced above: ±400 ppm/°C at 0.5–0.99 mΩ, ±275 at 1–2.9 mΩ, ±110 at 5–6.9 mΩ and ±75 at 50–100 mΩ.
- Vishay Dale, WSK2512 Power Metal Strip Resistors, Improved Stability. The four-terminal version: 0.0005 to 0.2 Ω, 1.0 W at 70 °C, ±35 ppm/°C from 5 mΩ up, and “4-terminal design allows for 1 % tolerance down to 0.0005 Ω” with separate sense and current connections.
- Isabellenhütte, MANGANIN data sheet. Resistivity 43 µΩ·cm at 20 °C, temperature coefficient ±10 ppm/K between +20 and +50 °C, thermal EMF against copper −1.0 µV/K, and a maximum working temperature in air of +140 °C with a recommendation to stay below +60 °C for the highest-precision use.