Last checked: 8 September 2026.

Every guide to testing a MOSFET with a multimeter ends the same way: charge the gate, watch the part turn on, declare it good. Almost none of them publish the two numbers that decide whether that test can work at all — the gate threshold the part is specified to, and the open-circuit voltage your meter’s diode range actually delivers.

Put those two numbers side by side across 17 datasheets and a whole class of perfectly good parts fails the test. The IRF530N, IRFZ44N and IRF3205 are specified to turn on anywhere up to 4.0 V of gate voltage. A Fluke 87V’s diode range tops out below 3.9 V. A $10 meter tops out around 3 V. The part is in spec, the meter is in spec, and the verdict is wrong.

The two numbers

The diode range is a current source with a voltage ceiling, and both figures are published. From the Fluke 83V and 87V detailed specifications, in the terminal characteristics table: diode function Open Circuit Test Voltage < 3.9 V dc, Full Scale Voltage 3.000 V dc, Typical Short Circuit Current 0.6 mA. From the manual of a DT-830L, the meter half the world owns: “Forward DCA is approx. 1mA”, “Open voltage is approx.3V”.

So the entire budget you have for turning a gate on with a handheld meter is roughly 3 to 3.9 V, at 0.6 to 1 mA. Nothing you do with the probes changes that. What changes from part to part is the other number.

Gate threshold across 17 parts

Every figure below is the manufacturer’s own, at 25 °C, read off the datasheet linked at the end. VGS(th) is the min–max window, not the typical value — the max column is the one that matters, because that is what the manufacturer allows a good part to need.

Part Type / package VDS VGS(th) min–max Spec’d at ID RDS(on) spec’d at VGS Turns on from a 3 V diode range?
AO3400A N, SOT-23 30 V 0.65 – 1.45 V 250 µA 2.5 / 4.5 / 10 V Yes
AON6414A N, DFN5x6 30 V 1.5 – 2.5 V 250 µA 4.5 / 10 V Yes, at the edge
AON6758 N, DFN5x6 30 V 1.4 – 2.4 V 250 µA 4.5 / 10 V Yes
CSD17573Q5B N, SON 5×6 30 V 1.1 – 1.8 V 250 µA 4.5 / 10 V Yes
CSD18540Q5B N, SON 5×6 60 V 1.5 – 2.3 V 250 µA 4.5 / 10 V Yes
PMV60EN N, SOT23 30 V 1.0 – 2.0 V 1 mA 10 V Yes
BUK9K6R2-40E Dual N, LFPAK56D 40 V 1.4 – 2.1 V 1 mA 5 V Yes
2N7002 N, SOT23 60 V 1.0 – 2.5 V 250 µA 10 V Yes, at the edge
IRLML6344 N, SOT-23 30 V 0.5 – 1.1 V 10 µA 2.5 / 4.5 V Yes
IRLZ44N N, TO-220 55 V 1.0 – 2.0 V 250 µA 5 V Yes
IRF530N N, TO-220 100 V 2.0 – 4.0 V 250 µA 10 V Not guaranteed
IRFZ44N N, TO-220 55 V 2.0 – 4.0 V 250 µA 10 V Not guaranteed
IRF3205 N, TO-220 55 V 2.0 – 4.0 V 250 µA 10 V Not guaranteed
AO3401A P, SOT-23 −30 V −0.5 – −1.3 V 250 µA −2.5 / −4.5 / −10 V Yes
AON6403 P, DFN5x6 −30 V −1.2 – −2.2 V 250 µA −4.5 / −10 V Yes
AON7423 P, DFN3.3×3.3 −20 V −0.2 – −0.9 V 250 µA −1.5 / −1.8 / −2.5 / −4.5 V Yes
PMV48XP P, SOT23 −20 V −0.75 – −1.25 V 250 µA −4.5 V Yes

The split is not random. Everything in the top block is a modern trench part designed to be driven by a 3.3 V or 5 V controller — the class you find on a laptop, console or phone board. The three parts marked not guaranteed are the older standard-level TO-220s from power supplies, amplifiers and motor drives, and they are also the parts most likely to be sitting in front of someone following a generic tutorial. That is why the test has a reputation for condemning good parts.

How to run the test with the table in front of you

Out of circuit, on a discrete N-channel part, in diode mode:

  1. Short all three legs together with a probe tip or tweezers first. This discharges the gate. Skipping it is the single most common reason two people get two different results on the same part.
  2. Body diode. Black probe on drain, red on source: you should read a forward junction. Reverse the probes: OL. If both directions read near zero and the meter beeps, the part is shorted and you are done.
  3. Charge the gate. Red probe to gate, black to source, for a second. Then move red back to drain, black stays on source. If the part turned on, the reading collapses toward zero.
  4. Discharge. Short gate to source again; the drain–source reading should go back to OL.

Step 3 is where the table earns its keep. Before you call step 3 a failure, look up VGS(th) max for that part number. If it is above roughly 3 V, your meter cannot be trusted to complete the test, and a “dead” verdict means nothing. Drive the gate from a bench supply at 10 V through a 10 kΩ resistor instead, then repeat step 3 with the meter.

The part nobody covers: the threshold is defined at your meter’s own test current

Look at the fifth column again. VGS(th) is specified at a drain current of 250 µA on most of these parts, 1 mA on the Nexperia devices, and 10 µA on the IRLML6344. Your meter’s diode range sources 0.6 mA to 1 mA.

Those are the same number. Which means that when a FET “turns on” during a meter test, it has reached exactly the condition the datasheet calls threshold — a few hundred microamps of drain current — and nothing beyond it. RDS(on) is a different specification at a different gate voltage: 4.5 V or 10 V in every row of the table, never 3 V. So the test proves the gate is intact and the channel forms. It proves nothing whatsoever about on-resistance, and a part whose RDS(on) has drifted up but still switches will pass it clean.

The second consequence is that the test current is not standardised. Comparing a 10 µA threshold to a 250 µA threshold to a 1 mA threshold is comparing three different measurements of three different things, which is worth knowing before you decide one part number is “lower threshold” than another. Infineon prints the proof inside a single document: the IRLML6344 datasheet plots threshold voltage against temperature with two curves on the same axes, one for ID = 10 µA and one for ID = 250 µA.

And it moves with temperature

Nexperia publishes the drift for the PMV60EN outright, at three junction temperatures:

Tj VGS(th) min VGS(th) max
25 °C 1.0 V 2.0 V
150 °C 0.6 V —
−55 °C — 2.2 V

Nexperia sells the same behaviour as a feature on the BUK9K6R2-40E: “True logic level gate with VGS(th) > 0.5 V @ 175 °C”. On the bench this cuts two ways. A part that measures fine cold will start conducting at a lower gate voltage once the board is hot — which is what an intermittent that only appears after ten minutes actually looks like. And testing cold with a marginal meter is the worst case for the test, not the best.

Limits and common errors

What you see What it means What to do
~0.4–0.6 V one way, OL the other, gate shorted to source Healthy body diode Go to the gate test
Near zero both ways, beeper on Hard short — part is dead Replace, and find what killed it before powering up
OL in both directions Open, or you have the pinout wrong Check the pinout in the datasheet before condemning
Gate charged, part still won’t conduct Dead gate or VGS(th) max above what your meter delivers Check the table; if max > 3 V, drive the gate externally
Turns on, then falls off by itself Gate not holding charge — in circuit, that’s the board’s gate pull-down Retest out of circuit before concluding anything
Body diode reads far below the datasheet figure Normal, not a fault See below

That last row catches people constantly. VSD in a datasheet is specified at rated current, not at meter current: 1 A for the AO3400A, 15 A for the BUK9K6R2-40E, 25 A for the IRFZ44N, 35 A for the CSD17573Q5B, 62 A for the IRF3205 — all quoted as 1.0 to 1.3 V. Your meter injects about 1 mA, three to five orders of magnitude lower, so you will read 0.4 to 0.6 V on a perfectly good junction. The same effect on ordinary diodes, with the manufacturer’s own curves, is worked through in detail elsewhere on this site.

Two more limits worth stating plainly. In circuit, the gate is tied to a driver output and usually a pull-down resistor, so it will not hold charge and step 3 is not available to you — the body diode check still is. And on the Fluke, the diode range’s full-scale display is 3.000 V, so anything above that reads OL whether or not the junction is fine.

Where this sits in a repair

A FET is rarely the first thing you should be looking at. Let a current-limited bench supply narrow the fault to one rail first, then read the schematic and boardview to find out which regulator owns it — that post covers a controller that deliberately clamps a healthy rail to ground and looks exactly like a shorted FET. Know what your meter’s specifications actually promise before you trust a borderline reading, and remember that a dried-out filter capacitor and a leaky FET produce very similar symptoms. Getting the part off the board without cooking its neighbours is a hot air job.

FAQ

My IRF530 won’t turn on when I charge the gate. Is it dead?
Not necessarily, and this is the case the table exists for. Infineon specifies that part’s threshold as high as 4.0 V; no handheld meter’s diode range guarantees that much. Drive the gate from a supply at 10 V through 10 kΩ and retest.

The body diode reads 0.45 V, but the datasheet says 1.3 V. Is it leaky?
No. The datasheet figure is measured at tens of amps. At the ~1 mA your meter injects, 0.4–0.6 V is what a healthy silicon junction reads.

Does passing this test mean the FET is good?
It means the gate works and the body diode is intact. It does not mean RDS(on) is in spec, because on-resistance is specified at 4.5 V or 10 V of gate drive and this test never gets near that. A part that has drifted high but still switches passes.

Sources

No bench measurements are reported here. Every figure above is the manufacturer’s published specification, taken from the datasheets linked in this list.