How to Test a Proximity Sensor With a Multimeter

Three measurements that settle the question

A proximity sensor is judged on three measurements: supply voltage at the sensor itself, current draw, and the voltage on the signal output when metal enters the zone. That is enough to tell a dead sensor from a live one and to settle the output type when the housing marking has worn off.

One condition governs the whole result: the output must be measured with a load. On an unloaded output a multimeter shows not the state of the sensor but the consequence of its own input impedance, and because of that good items are regularly written off as faulty. Below is the order of work, the normal limits from official datasheets, and a table mapping a reading to its cause. Every figure comes from the manufacturer's datasheet for a specific model rather than from a generic table.

Safety: what to do before you touch the terminals

First establish the voltage in the circuit

In the three-wire DC models this article refers to, the supply range is 10…30 V. That is not a universal figure: the P+F NJ2-12GM40-E2-V1 datasheet states UB 10…60 V, so three-wire versions for a wider range do exist. Take the range from the datasheet for your own item rather than from the assumption that a sensor means 24 V. Two-wire types can be different again. For Pepperl+Fuchs NBB20-U1-UU the datasheet gives a rated operating voltage Ue of 24…250 V DC or 24…230 V AC, and an operating voltage range UB of 20…275 V DC or 20…253 V AC. The bottom of that range is 24 V DC, so a 24 V DC laboratory supply tests this sensor within its own specification: 24 V sits inside the rated values.

The hazard lies elsewhere. The same sensor in a cabinet may be sitting at 230 V AC. If you do not know the circuit voltage for certain, treat it as mains: de-energise the cabinet, confirm the absence of voltage with a voltage indicator, remove the sensor and test it on the bench from a 24 V DC laboratory supply. Testing a two-wire sensor in place in the cabinet at mains potential is not part of this procedure, and this article does not describe it.

What the instrument has to be

CAT III 300 V is the minimum, and it is rated for a 230 V circuit. If the cabinet the sensor came out of is fed from a 400 V three-phase supply, the instrument needs to be CAT III 600 V. Probes of the same category as the instrument, insulation with no cracks and no scorch marks. A giveaway lead set rated 600 mV with no category marking is not suitable for this work, even if you only intend to measure 24 V: two-wire sensors that came out of a cabinet turn up in the same boxes as 24 V three-wire ones.

Required modes: V DC, mA, continuity and diode test. Plus a 24 V DC source. Plus a load: a 2.2 kΩ 0.5 W resistor, or a 24 V relay, or a 24 V lamp.

The mA range: what exactly makes it dangerous

In mA mode the probes sit across the meter's internal shunt, and that shunt has a resistance of the order of single ohms or less. In this mode the instrument is only good for insertion into a break in the circuit, in series with the load. If, with the probes in the mA or A jack, you touch two points in parallel with the source, you short that source through the shunt. On a 24 V bench this ends with the mA jack's fuse blown, and in most instruments replacing that fuse means opening the case. In a mains circuit the same touch produces an arc flash, and the consequences are no longer about the instrument.

Hence three rules. First: in mA mode the instrument goes into a break only, never in parallel. Second: circuits are de-energised before probes are moved, and power is restored after the probes are secured. Third: in a mains two-wire circuit current is not measured with a multimeter at all, because probes in the mA jack next to mains potential is an accident with a predictable outcome. After a current measurement, return the probes to the V jack immediately.

Two datasheet constraints that apply to mains-rated two-wire types and are easy to break on a bench. First: the NBB20-U1-UU datasheet explicitly requires an external fuse in accordance with UL 248, rated 3 A and suitable for 300 V AC and DC. Second: the same datasheet states that the sensor is designed for individual operation and that a series or parallel connection of sensors is not permissible. So "I will test two at once in the same circuit" is not a time saving, it is outside the specification.

NO or NC: a caveat that applies to every table below

Every table in this article is written for a normally open output (NO): at rest the output does not conduct, and it conducts when metal enters the zone. For a normally closed (NC) output every line inverts, and what the NO table calls a fault turns out to be correct operation.

So the switching function has to be known before the measurements, not after. It comes from the datasheet or from the part number. There is a separate class of cases where neither settles it: the NBB20-U1-UU datasheet states the switching function as "Normally open/closed (NO/NC) selectable through wiring", that is, set by how the sensor is connected. Such a sensor has no correct reading in itself; it depends on how it was wired in the cabinet. Record the existing wiring before you test.

Step 1: supply voltage at the sensor

Measure V DC between the brown and blue wires, and do it at the far end of the sensor cable rather than at the cabinet terminal. The difference between those two points is itself a standard cause of support calls.

The normal range for three-wire models in our catalogue is 10…30 V. That is what both datasheets state: Sick IME12-04BPSZC0K (10 V DC … 30 V DC, ripple ≤ 10 %) and P+F NBB5-18GM50-E2 (UB 10…30 V).

What the readings mean:

• 0 V: a cable break, a tripped breaker, or a failed power supply output.
• Below 10 V: a voltage drop. Work through the causes in order, cheapest check first: conductor cross-section, run length, total load on the supply. A sensor in that state works intermittently, which presents as a floating fault, and people end up swapping first the sensor and then the controller while the power supply is at fault.
• 24 V at the terminal and 0 V at the sensor: the break is in the cable, most often where it flexes on a moving part.
• Ripple above the rated 10 %: a meter in V DC mode will not show this, only the average. If the supply is unstable and the cause is not found, that is work for an oscilloscope.

Step 2: supply current I0

If the voltage is fine but the sensor stays silent, measure the current. Switch the instrument to mA, break the brown wire and insert the meter into the break.

An important detail that is easy to miss: the rated I0 is specified without a load. The full parameter name is "No-load supply current" at P+F and "No load current" at Sick. So the load must be taken off the output for the duration of this measurement; otherwise you get the sum of the supply current and the load current, and comparing that against the datasheet means nothing.

Normal means up to the rated I0 of your own model, not an abstract "a few tens of milliamps". The figures differ:

So 25 mA is already out of specification for NBB5-18GM50-E2, while for IME12-04BPSZC0K out of specification starts above 10 mA. Read the datasheet for your item, not for the one next to it.

• Near zero: the sensor is not being powered although voltage is present at the terminals. Check polarity, brown to plus and blue to minus. The NBB5-18GM50-E2 datasheet declares reverse polarity protection, so swapped wires will not kill it, but it will not work either.
• Within the rated I0: the electronics are alive, so continue.
• Above the rated I0: a short on the output or failed electronics. There is no point testing such an item further.

Step 3: NPN or PNP

This step needs a load. Wire the 2.2 kΩ resistor as shown in the table and measure the voltage between the black and blue wires. In both cases the probes sit on the same two points, black and blue; only the resistor's connection changes.

This table is for NO. For NC the bottom two rows swap.

On the expected values in the "metal in the zone" state. The voltage drop across a conducting output is in the datasheet, but it is worded differently by each maker. Sick states "Voltage drop ≤ 2 V" for IME12-04BPSZC0K with an explicit footnote "At Ia max", and Ia there is ≤ 200 mA, so the limit is tied to maximum current openly. P+F states Ud ≤ 3 V for NBB5-18GM50-E2 with no current condition attached, and that is simply a limiting value. The current in your own circuit equals the reading divided by the resistor value: 2.2 kΩ with an ideal zero drop would give 10.9 mA. For a PNP type with metal present, expect about 23.5…24 V, because at around 10 mA the real drop is a fraction of a volt. For comparison: the P+F NJ2-12GM40-E2-V1 datasheet lists Ud 1.2…2.5 V at IL = 100 mA alongside the limiting Ud ≤ 3 V, so even at ten times the current the drop does not reach the limit.

On the "no metal in the zone" state. For PNP this is not strictly zero. An off-state current Ir flows through the non-conducting output, and across a 2.2 kΩ resistor the rated maximum of 0.5 mA would give 1.1 V. So up to roughly 1.1 V on a loaded PNP output that is off is still normal. The typical Ir of 0.1 µA at 25 °C gives less than a millivolt across the same resistor, so on a healthy sensor you will read nearly zero.

A reading of 21 V with metal present is formally still inside the rated Ud ≤ 3 V. But the current here is 21 V divided by 2.2 kΩ, about 9.5 mA, and a 3 V drop at that current is anomalous. It is not "normal per the datasheet" but a reason to test the sensor under its real load by step 4.

Dissipation in the resistor is 24²/2200 ≈ 0.26 W, which is why it should be rated 0.5 W.

A 24 V lamp or a 24 V relay will do instead of the resistor. A relay gives an audible click, so you get confirmation by ear and your hands stay free. A lamp gives confirmation by eye. In both cases check the coil or lamp current against the rated output current: IL 0…200 mA for NBB5-18GM50-E2, Ia ≤ 200 mA for the Sick IME12 types.

Why a voltmeter lies on an unloaded output: two different mechanisms

In V DC mode a multimeter presents an input impedance in the order of 10 MΩ. To a sensor output that is practically an open circuit. From there, though, PNP and NPN behave differently and for different reasons. The two get confused, and wrong conclusions follow.

PNP: off-state current Ir

A PNP output switches the positive rail, so the load belongs between black and blue. If a voltmeter stands in place of the load, all that remains in the circuit is its 10 MΩ.

A transistor that is off is not a perfect insulator: an off-state current flows through it, and the datasheet specifies it. For NBB5-18GM50-E2 that is Ir 0…0.5 mA, typically 0.1 µA at 25 °C. Work out what that current does across 10 MΩ. A typical 0.1 µA gives 1 V. To push the reading to a full 24 V, 2.4 µA is enough, and that is two hundred times less than the rated maximum of 0.5 mA. So an unloaded PNP output that is off will predictably read up to the full supply voltage on a voltmeter, and that is a property of the measurement rather than a fault in the sensor.

The practical conclusion: 24 V on an unloaded PNP output at rest proves nothing. The diagnosis is made only with a load.

NPN: induced voltage

An NPN output switches the negative rail, so the load belongs between black and brown. Without a load the black wire is pulled neither to plus nor to minus: it floats.

Here the mechanism is different. No current leaves the sensor at all, yet there is a reading, because a free wire acts as one plate of a capacitor with respect to the neighbouring cores and the cabinet body. The instrument reads induced voltage, and that depends on how the cable runs, what is switching nearby and whether a variable frequency drive is working in the same cabinet. The reading can be arbitrarily unstable.

Why the two are not the same thing

Off-state current Ir is a rated parameter of the sensor: a real current through the non-conducting output, which across a high resistance produces a high voltage. Induced voltage is not a sensor parameter at all: it is a property of a floating conductor in an environment full of switching, and no datasheet contains it or could.

The distinction has practical weight. Ir is dealt with by choosing a load and comparing against the input threshold. Induced voltage is dealt with by tying the conductor down, shielding, and separating cable runs. Mistake one for the other and you treat the wrong illness.

The same Ir also causes false triggering at a PLC input, but there in the working circuit. For the three-wire PNP NBB5-18GM50-E2 it is tiny: typically 0.1 µA at 25 °C, 0.5 mA maximum. For the two-wire NBB20-U1-UU the datasheet states Ir ≤ 1.5 mA, which can already be enough for a controller input to read a non-conducting sensor as switched on.

Step 4: switching under a real load

A resistor confirms the logic but not the ability to drive a real load. Operating current is specified separately: IL 0…200 mA for NBB5-18GM50-E2, continuous current Ia ≤ 200 mA for Sick IME12-04BPSZC0K.

Replace the resistor with a 24 V relay or lamp and pass a metal object through the zone ten times. If the relay does not click every time, there are usually two causes: the sensor output has degraded, or the object passes at the edge of the sensing zone.

On that edge, keep the datasheet figures in mind. IME12-04BPSZC0K has a nominal zone Sn of 4 mm and an assured zone Sa of only 3.24 mm. Layouts are designed to Sa. The non-flush IME12-08NPSZC0S in the same M12 diameter is rated 8 mm nominal with an assured zone of 6.48 mm. Add the rated temperature drift of ±10 % of Sr and it becomes clear why a layout set exactly to Sn works on the bench and fails on the shop floor in summer.

Two-wire sensors: a different procedure

A two-wire sensor goes in series with its load, like a limit switch. The current that powers it passes through the load, and everything specific about it follows from that.

Why a voltmeter in series reads like an open circuit

Put a voltmeter in series with a two-wire sensor that is off and the instrument shows almost the full supply voltage. People read that as a break inside the sensor and write it off.

In fact that is exactly what should happen. The NBB20-U1-UU datasheet lists Im, the lowest operating current, as 5 mA. A voltmeter of 10 MΩ passes microamps, three orders of magnitude short of that, so the sensor never switches on in such a circuit. Meanwhile a sensor that is off passes an off-state current of up to 1.5 mA, and across 10 MΩ that gives a voltage limited only by the supply. The result: almost the full supply on the display, whether or not metal is in the zone.

So a two-wire sensor is not tested with a voltmeter in series. No reading in that circuit is a diagnosis.

How to test it properly

Build a circuit of plus, sensor, load, minus, where the load draws at least Im, and measure the voltage across the load.

Let us size the load for NBB20-U1-UU on a 24 V DC bench. At least 5 mA is needed. In the on state the sensor itself takes the drop Ud, rated at up to 5 V on DC (typically 2 V), so about 19 V is left across the load in the worst case. Hence the upper resistance limit: 19 V divided by 5 mA gives 3.8 kΩ, so the nearest standard value is 3.3 kΩ. The lower limit is set by dissipation: 1.2 kΩ at 24 V gives 20 mA and 0.48 W, which is already at the edge for a 0.5 W resistor, so below 1.2 kΩ use a 1 W part.

The working choice: 2.2 kΩ rated 0.5 W. Across 19 V that is 8.6 mA, comfortably above Im of 5 mA, dissipating 0.16 W. A 4.7 kΩ value gives only 4 mA in this calculation and the sensor may fail to switch on, so it is not worth using even though at first glance it looks suitable.

What to expect on the measurement: in the on state about 19…22 V across the load and the drop Ud across the sensor. The NBB20-U1-UU datasheet gives Ud ≤ 5 V (typically 2 V) on DC and ≤ 8 V (typically 3 V) on AC. The difference between DC and AC matters here and must not be carried from one mode to the other.

When a PLC input latches on

The classic situation: the two-wire sensor is off and the controller input still reads a one. The cause is the off-state current, 1.5 mA for NBB20-U1-UU, and that current across the PLC's input impedance produces a voltage above the off-state threshold.

Two solutions, both standard practice. First: a bleeder resistor in parallel with the PLC input. It diverts the off-state current and the input voltage falls below the threshold. The value is calculated from the controller manual: take the guaranteed off-state input voltage and divide it by the sensor's rated Ir, and that is the upper resistance limit. We cannot quote figures here, because the threshold differs between PLCs and has to come from the manual for yours.

Second, and the more robust answer: a three-wire sensor instead of a two-wire one. A three-wire PNP has an off-state current two or three orders of magnitude lower, typically 0.1 µA at 25 °C for NBB5-18GM50-E2, so the problem does not arise in the first place.

One limit on this section. The comparison with a 24 V PLC input holds only for DC. If the two-wire sensor sits in an AC circuit, that is a different arrangement, with different thresholds and a different Ud (≤ 8 V, typically 3 V), and the logic of a 24 V DC digital input does not apply to it.

What a multimeter can check, and what it cannot

The ohmmeter and the diode range

Putting an ohmmeter across a modern sensor's terminals in the hope of seeing "coil resistance" is pointless: inside there is an oscillator with an output transistor, not a coil brought out to the terminals.

The diode range, though, gives one useful result. The supply input of a three-wire sensor carries reverse polarity protection: the NBB5-18GM50-E2 datasheet has the line "reverse polarity protected", NBB20-U1-UU has "reverse polarity tolerant". So in diode mode between brown and blue you will most likely see a junction reading one way round and no conduction the other. The presence of that asymmetry is not a fault; it is the protection device showing itself.

What is a diagnosis: a dead short between brown and blue, that is, near zero ohms and the instrument beeping in both directions. That item is dead and needs no further testing.

Away from the sensor, the ohmmeter and the continuity range are useful on the cable, and the cable is where to start when a sensor "works sometimes".

What cannot be measured at all

• Switching frequency. IME12-04BPSZC0K is rated at 2,000 Hz, NBB5-18GM50-E2 up to 800 Hz, the two-wire NBB20-U1-UU 0…100 Hz on DC and 0…30 Hz on AC. Confirming that takes an oscilloscope or a fast controller input, not a multimeter.
• Time delay before availability. For IME12-04BPSZC0K it is ≤ 100 ms, for NBB20-U1-UU tv ≤ 300 ms. A multimeter does not resolve such intervals, and in practice they mean the first readings after power-up are not worth reading.
• The real sensing zone on your metal. The rated Sn is given against steel. Reduction factors appear in every datasheet and they differ even between neighbouring models. NBB5-18GM50-E2 states rAl 0.3, rCu 0.3, r304 0.7. NBB20-U1-UU states rAl 0.4, rCu 0.3, r304 0.77, rBrass 0.43, rFe 1. Sick IME12-04BPSZC0K states St37 1, stainless V2A/304 approx. 0.8, aluminium approx. 0.45, copper approx. 0.4, brass approx. 0.4, and the datasheet itself adds that these are reference values which may vary. No single table covers all sensors, and a factor from another model cannot be substituted.

When the marking misleads

A part number can only be read against the manufacturer's ordering table; reading it "by the meaning of the letters" produces errors. A working example from our own catalogue: Sick IME12-08NPSZC0S. The code contains NP, and that is regularly read as NPN. Datasheet 1040780 says otherwise: Switching output PNP, Output function NO, Installation type Non-flush. So N here stands for non-flush mounting and P stands for PNP, and the letters belong to different fields of the code.

Hence the order of work. If there is a datasheet for the part number, take the output type from the datasheet. If there is no datasheet or the part number does not parse, measure it by step 3 with a resistor. Guessing from the letters in the code is the most expensive way to be wrong.

Wire colours: what you can rely on

The convention of brown for plus, blue for minus and black for the output is guaranteed for sensors that declare conformity with EN IEC 60947-5-2. The NBB5-18GM50-E2 and NBB20-U1-UU datasheets declare that conformity explicitly. On a cable without it, colours have to be checked rather than assumed.

Two separate cases where the step 3 table does not behave as expected:

• Four-wire versions. There is a second signal wire, usually white, and it is either a second switching function (antivalent NO and NC at once) or a second output. Both have to be measured, one after the other, because one of them always conducts at rest.
• Push-pull outputs. Such an output both sources from plus and sinks to minus. The step 3 table will give a "correct" result in either resistor arrangement, so it cannot distinguish push-pull from a plain PNP or NPN. Here the datasheet or ordering table is required.

Reading-to-cause table

For NO outputs. For NC, invert the rows that concern the output state.

Frequently asked questions

Can NPN and PNP be told apart without a resistor?

No, and this is the main correction to the advice usually given. Without a load, a PNP that is off reads almost the full supply because of its off-state current, and an NPN that is off reads induced voltage. Neither reading depends on the state of the sensor, so the output type cannot be deduced from them. A 2.2 kΩ resistor removes the uncertainty completely.

The meter reads 24 V on the output constantly and metal makes no difference

First check whether there was a load. If you measured an unloaded PNP output, 24 V at rest is normal rather than a diagnosis: that is Ir working through the instrument's 10 MΩ. Repeat with a 2.2 kΩ resistor between black and blue.

The verdict "the output does not turn off" applies only when 24 V persists under load both at rest and with metal in the zone. Then the output transistor has failed and the sensor gets replaced.

In the NPN arrangement the resistor goes to brown, and the voltage sits constantly near 24 V

The arrangement is right, and 24 V there means the output is not conducting. Three causes, in order of likelihood. First: the metal never reaches the sensing zone, that is, the gap exceeds Sa or the object's material has a reduction factor you did not allow for. Second: the output has degraded and no longer pulls the black wire down to zero. Third: the sensor turned out to be NC and metal is permanently in the zone, from a bracket or a neighbouring part, in which case 24 V is the normal "operated" state.

Separately, about an error that sounds logical but is physically impossible: if in the NPN arrangement you connect the resistor between black and blue, you will never see 24 V. In that circuit the black wire is tied only to zero and to nothing else, so the reading will be close to 0 V in both states of the sensor. A constant zero with the resistor on blue indicates a wrong test circuit, not a faulty sensor.

Which resistor should I use for the test?

For three-wire types on 24 V: 2.2 kΩ rated 0.5 W. That is about 10.9 mA, enough current for a clean reading and well inside the rated 200 mA.

For the two-wire NBB20-U1-UU on 24 V DC: the same 2.2 kΩ at 0.5 W, because with the rated Ud of up to 5 V it gives 8.6 mA, comfortably above Im of 5 mA. The workable range here is roughly 1.2…3.3 kΩ, and below 1.2 kΩ a 1 W part is needed because dissipation approaches 0.48 W.

A 24 V lamp or a 24 V relay works instead of the resistor. The relay is the more convenient of the two, because the click gives confirmation by ear and leaves your hands free.

The sensor runs warm. Is that normal?

Mild warmth on the housing happens at currents close to the maximum. Noticeable heat together with a current above the rated I0 means a short in the load or a failed output. Check the rated temperature limits as well: NBB5-18GM50-E2 is rated −25…70 °C, Sick IME12-04BPSZC0K −25…+75 °C, NBB20-U1-UU −25…85 °C.

Should the sensor be disconnected from the PLC during testing?

Yes. Disconnect the signal wire from the input, otherwise you are measuring in parallel with the controller's input circuit and reading the sum of two paths. The same applies to step 2: I0 is specified without a load, so the load comes off for the duration of the measurement.

What if the sensor is two-wire and sits in a cabinet at 230 V?

De-energise the cabinet, confirm the absence of voltage with an indicator, remove the sensor and test it on the bench from a 24 V DC laboratory supply with a load in the circuit. For NBB20-U1-UU that is a rated condition: its Ue starts at 24 V DC. Measuring current with a multimeter in a mains circuit is neither necessary nor safe.

The sensor tests good but the line still sees no signal

Then the mismatch is on the controller side: the input type does not match the sensor output type. The procedure is set out in connecting a sensor to a PLC. If the signal feeds a variable frequency drive, see replacing a European VFD with Veichi and the PLC jumper.

Do these steps work on a photoelectric or capacitive sensor?

Steps 1, 2 and 3 carry over, because the logic of a three-wire output is the same. The normal limits do not: I0, Ud, operating current and Ir come from the datasheet for your own model. In photoelectrics the sensing range is governed by different parameters, and metal reduction factors do not apply to it.

If the sensor is dead

Once all four steps are done and the item really does not work, send us through our contacts the part number and what the multimeter showed at each step. We will check your delivery note to see whether the item is still inside its warranty period, and pick a replacement in parallel. Warranty periods we read off the supply documents; we do not quote them from memory. If you did not buy the sensor from us we cannot check its warranty, because we hold no supply documents for it, but we will work out a replacement from your measurements just the same.

A replacement is chosen on three parameters: housing diameter, assured zone Sa, output type. The selection principles are set out in inductive sensors: operating principle and selection, and the procedure for when the original is no longer supplied is covered separately in our piece on finding an equivalent.

A guide by housing size, to narrow the search:

Availability and lead time on each item we confirm on request, because stock moves daily.

One detail people forget when replacing a sensor. The sensor ships without a cable, so it needs one with the mating socket. An M12 connector is fitted on every item in the table above except IME08-02BPSZT0K, which has M8. Any of these M12 sensors takes P+F V1-W-5M-PUR: an M12 A-coded angled 4-pin socket, 5 m PUR cable, up to 250 V AC/DC and 4 A, IP66/IP68/IP69. The three-pin M12 versions (the newer revisions of NBN8-18GM50-E2-V1 and NBB8-18GM50-E2-V1) use pins 1, 3 and 4 of that same connector, so a four-pin socket mates with no modification. A separate cable is needed only for the M8 3-pin on IME08-02BPSZT0K: that is a different thread size, and we pick it on request. If the old sensor already had a cable and it is intact, a new one is not needed, and step 1 shows that: the supply at the far end of the cable reads normal.

Catalogue: Sick sensors, Pepperl+Fuchs sensors, Omron sensors, all industrial sensors. We stock two-wire DC types as well, for example NBB5-18GM40-Z0-V1; take its current limits from its own datasheet, because among two-wire types they vary more than among three-wire ones.

Send us through our contacts four things: housing diameter, the distance you need, the object material, the controller input type. We will check it against the manufacturer's datasheet and offer an item.