Specification for Testing the On-resistance of Discrete Semiconductor Devices

On-Resistance Testing for Discrete Semiconductors: The Specs People Actually Follow

Reading the on-resistance spec on a datasheet is easy. Measuring it correctly is a different story. Rds(on) for a MOSFET, Vce(sat) for a BJT, Vf at a given current for a diode — these numbers look simple until you realize that temperature, current level, gate drive voltage, and even the test pulse width all shift the result. A part that reads 10 milliohms on one tester might read 14 milliohms on another, and both readings are correct. The question is which one matches your actual operating condition.

What On-Resistance Actually Tells You

On-resistance is the resistance of the device when it is fully turned on. For a MOSFET, this is the drain-to-source resistance with the gate driven hard into conduction. For a BJT, it is the collector-to-emitter saturation voltage divided by the collector current. For a diode, it is the forward voltage drop divided by the forward current at a specified operating point.

The reason this matters is that on-resistance directly determines power loss in the device. A MOSFET with 20 milliohms Rds(on) carrying 10 amps dissipates 2 watts. Drop that to 10 milliohms and you are at 1 watt. That difference decides whether you need a heatsink or not.

But here is the catch: the on-resistance spec on the datasheet is measured at a specific gate voltage, a specific drain current, and a specific junction temperature — usually 25°C. Your circuit runs at a different temperature, a different current, and possibly a different gate drive. The datasheet number is a reference point, not a guarantee of what you will see in the wild.

How to Set Up a Valid On-Resistance Measurement

Four-Wire Kelvin Sensing Is Mandatory

Two-wire measurements include the resistance of the test leads, the contact resistance at the probe points, and any solder joint resistance in the path. For a device with an on-resistance in the milliohm range, the lead resistance alone can be a significant fraction of the total. A two-wire measurement of a 5 milliohm MOSFET with 0.2 ohms of lead resistance gives you a reading that is 40 times too high.

Use four-wire Kelvin connections. Two wires carry the current, two separate wires sense the voltage directly at the device terminals. This eliminates lead resistance from the measurement entirely. Most benchtop meters and SMUs have Kelvin sense terminals built in. If yours does not, you need external sense leads clipped as close to the device pins as possible.

Current Level Must Match the Application

On-resistance is not a constant — it varies with current. A MOSFET might show 8 milliohms at 1 amp and 12 milliohms at 50 amps because the channel does not conduct uniformly at high current densities. If you measure at 100mA and your circuit runs at 20 amps, your measurement is useless.

Always set the test current to match the actual operating current in your circuit. If the part will see a range of currents, measure at the highest current you expect, because that is where the losses are worst and where the thermal stress is greatest.

Temperature Control: The Part Nobody Gets Right

Junction Temperature vs. Ambient Temperature

On-resistance increases with temperature. For silicon MOSFETs, the temperature coefficient is typically positive — Rds(on) can be 1.5 to 2 times higher at 125°C than at 25°C. For some technologies, the coefficient can even be negative at low currents and positive at high currents, which makes things messy.

The datasheet spec is almost always given at 25°C junction temperature. But in a real board, the junction is at 85°C or 125°C. If you measure on-resistance at room temperature and multiply by a derating factor, you are guessing. The proper way is to heat the device to the target junction temperature, then measure.

A hot chuck or a temperature-controlled fixture under the device works well. Let the part soak at temperature for at least two minutes before measuring — the silicon needs time to equalize thermally. If you measure too soon, the junction is still cooler than the case and your reading will be falsely low.

Pulse Testing to Avoid Self-Heating

When you push high current through a device to measure on-resistance, the device heats up from the inside. Even a 10-millisecond pulse at 50 amps can raise the junction temperature by several degrees in a small package. That self-heating lowers the on-resistance reading because you are measuring at a higher temperature than you think.

Use short pulses — 100 microseconds to 1 millisecond — with a low duty cycle. This gives enough time for the current to stabilize and the voltage to settle, but not enough time for the junction to heat up significantly. Most SMUs and curve tracers have a pulse mode specifically for this. Set the pulse width short, trigger the measurement at the end of the pulse when the reading has settled, and wait at least 100 milliseconds between pulses to let the device cool.

Gate Drive Voltage Changes Everything for MOSFETs

Underdriving the Gate Gives Fake Numbers

The on-resistance spec on a MOSFET datasheet is given at a specific Vgs — usually 10V or 4.5V depending on the device. If you drive the gate at 3.3V instead of 10V, the channel does not form fully and the on-resistance can be three to five times higher than the spec.

When testing, use the exact gate voltage your circuit will use. If your driver runs at 3.3V, measure at 3.3V. Do not measure at 10V and assume it will scale linearly — it does not. The transfer curve is exponential in the subthreshold region, so small changes in gate voltage near the threshold cause huge changes in on-resistance.

Gate Charge Affects the Transition

For switching applications, the on-resistance during the transition matters as much as the steady-state value. When the gate voltage is ramping up, the device passes through a region where it is partially on — high resistance, high dissipation. This is where most switching losses happen.

A standard DC on-resistance measurement does not capture this. To characterize switching behavior, you need to look at the voltage and current waveforms during turn-on and turn-off with an oscilloscope. The area under the V times I curve during the transition gives you the switching loss per cycle. Multiply by frequency and you have your total switching power dissipation.

Interpreting the Data When It Does Not Match the Datasheet

Normal Variation vs. Real Failure

Datasheet on-resistance values come with a range. A part spec'd at 10 milliohms max might actually measure anywhere from 6 to 10 milliohms on a good day. If your reading is 11 milliohms, the part is not necessarily bad — it might just be at the upper end of the distribution.

Compare against the maximum spec, not the typical value. If the reading is under the maximum at your test conditions, the part passes. If you need tighter matching — say for current-sharing in parallel devices — you need to bin the parts yourself by measuring each one and sorting them into groups.

Contact Resistance Is Hiding in Your Measurement

If your on-resistance reading is higher than expected but the device tests fine on a curve tracer, the problem is almost certainly in your test setup. Poor probe contact, oxidized leads, or cold solder joints add resistance that shows up in your measurement but is not part of the device.

Clean the device leads with isopropyl alcohol before testing. Use spring-loaded probes or clips with firm contact. If you are testing on a board, desolder one leg and measure the device out of circuit — this eliminates any parallel paths through the PCB that could skew the reading.

Test Sequence That Gives Repeatable Results

Start with the device at room temperature. Apply a short preconditioning pulse at low current to stabilize the junction. Then apply your test current at the target gate voltage, wait for the reading to settle, and record the voltage. Repeat three times and average the results. If the three readings vary by more than 5 percent, your contact is unstable — clean the leads and try again.

After the room temperature measurement, heat the device to the target junction temperature, soak for two minutes, and repeat the entire sequence. The ratio of the hot reading to the cold reading is your temperature coefficient, and that number is far more useful than either reading alone because it lets you predict on-resistance at any temperature within the operating range.