# The phase sensor and reflective marker: how the instrument measures phase

> Two accelerometers on the bearing housings tell you only one thing: how hard something is shaking. They can't tell one shaft revolution from the next, because their signal carries no reference to the angle of rotation. That reference comes from a third sensor, a laser one, which looks at a reflective marker and outputs one pulse per revolution. Below we cover what the instrument does with that pulse, where to place the marker, how to aim the sensor, and how to check in two minutes that the count isn't lying to you.

**In short:** The phase sensor produces one short pulse for every rotor revolution. From the interval between pulses, the instrument calculates rotation speed, and from the timing of the pulse it measures the phase of the vibration's running-speed component: the angular delay between the marker and the peak of the 1x sine wave. Without that reference point, there's nothing to extract 1x from and nothing to measure degrees from, so all you're left with is overall vibration as a single number, and no correction weight can be calculated. The marker sets the zero point for phase, but not the zero point for the correction angle: the weight's mounting angle is measured from the trial weight's location instead.

Source: https://axiline.pt/en/articles/the-phase-sensor-reflective-marker/  
Publisher: AXILINE · Vila Nova de Gaia, Portugal · +351 931 831 229 · axilinegeral@gmail.com

## What the instrument does with the phase sensor's pulse

The sensor does one simple thing: every time the reflective marker passes the beam, it fires off a short pulse. One pulse per revolution. From the interval between consecutive pulses, the software calculates rotation speed. What comes next is the whole reason this sensor exists.

Knowing when each pulse arrives, the software slices the continuous vibration signal into segments exactly one revolution long and adds them together: N consecutive revolutions, sample against sample. Anything that repeats in step with the rotation builds up through that addition. Anything unrelated to the rotation - bearing noise, a neighbouring pump, 50 Hz mains interference, random knocks - averages out. This is synchronous averaging, and without a reference pulse it's simply impossible: there's no way to know where each period starts.

From the curve averaged over N revolutions, harmonic analysis pulls out two numbers per channel: the amplitude of the 1x running-speed component - the part of the vibration at the rotation frequency, which is exactly what imbalance produces - and its phase. Phase (fields F1 and F2 in the interface) is the angular delay between the marker's pulse and the peak of the 1x sine wave, running from 0 to 360°. The Balanset-1A measures it to an accuracy of about ±1°, and works across an RPM range of 100-100,000.

Disconnect the phase sensor, and the picture is cut in half. In 'Vibrometer' mode the instrument will keep showing overall RMS vibration velocity (root-mean-square, fields V1s and V2s), because a broadband number doesn't need a reference point. But the 1x and phase fields stay empty.

Now for the calculation itself. The running-speed component is a vector: its amplitude in mm/s sets its length, its phase sets its direction. The instrument stores the vector before the trial weight, stores it again after, and divides the difference by the trial weight's known unbalance. That gives an influence coefficient - this particular 'rotor-supports-foundation' system's response to a mass at a known point. From that, the software calculates the correction's mass and angle. Take away the phase, and there's nothing left to subtract: two numbers with no direction don't form a vector difference.

- RPM. Without an accurate rotation speed, you won't find 1x in the spectrum, and you'll confuse it with mains interference: on a motor at 2970 RPM, the running-speed component sits at 49.5 Hz, right next to the mains frequency at 50 Hz.
- 1x amplitude. Balancing only reduces this part of the overall vibration.
- 1x phase. The vector's direction, without which no influence coefficient can be calculated.
- Synchronous averaging. Suppression of everything unrelated to the rotation, which is exactly why 1x stays visible even on a noisy machine.
- Order-referenced spectrum. The instrument marks the peak at the rotation frequency from an actual measurement, not from your guess.

> Balancing methods without a phase sensor do exist: the trial weight is moved sequentially between marked positions, and the problem is solved from amplitudes alone. They work. But every position means a separate run, and on a live production line, a run costs you the operator's sign-off, a run-up, and a coast-down. A phase sensor buys those runs back.

## The reflective marker: location, size, surface preparation

The marker is a consumable that costs almost nothing, and the entire measurement depends on it. Mistakes with it don't make things 'a bit worse' - they produce meaningless numbers, from which the instrument will faithfully calculate the wrong weight.

1. **Choose a spot that rotates 1:1 with the rotor** — The shaft, the impeller hub, a half-coupling, a spindle end face. The one requirement: that element has to make exactly one revolution for every one revolution of the rotor being balanced. The spot needs to be reachable by the beam with the guard closed, and clear of where the correction weight will eventually go.
2. **Make sure there's exactly one marker** — Walk the rotor and remove old stickers, scraps of tape, and bright patches of paint. Check for shiny details: a bolt head, a key, a chrome nut, a polished spot. Any one of them can produce a second pulse per revolution, doubling the reported RPM. On a thin shaft, watch that the tape doesn't wrap all the way around into a closed ring: at a 20 mm diameter, the circumference is only 63 mm.
3. **Size the marker to the peripheral speed** — Working guide: 10-20 mm around the circumference and 8-15 mm lengthwise. The pulse duration equals the marker's length divided by the peripheral speed. On a Ø100 mm shaft at 1500 RPM, that speed is 7.9 m/s, and a 10 mm marker passes by in 1.3 ms. On a Ø300 mm hub at 3000 RPM, it's already 47 m/s, and the same 10 mm gives you 0.2 ms. The faster the rotor, the wider you should make the marker, and the closer you should mount the sensor.
4. **Degrease and roll it down** — Wipe the pad with solvent or isopropyl alcohol, let it dry, apply the marker, and press it down for 10-20 seconds. Pay special attention to the leading edge: that's the part that peels first, from centrifugal force and airflow. Tape won't hold on flaking paint or rust - clean the strip down to bare metal.
5. **Get your contrast from reflection, not colour** — The sensor responds to reflected light returning to it, not to whiteness. On a polished, shiny shaft, the tape barely stands out - the whole surface bounces the beam back. First darken a ring around the marker with matte black paint or matte tape, then apply the reflective strip on top.
6. **Trace the edge with a marker pen** — The sensor triggers on the edge - that is, on the marker's boundary. Draw a thin line along the leading edge and note where it is. If the tape comes loose partway through the job, you can restore the reference point in exactly the same place and won't lose the phase readings you've already taken.

> Apply, clean, or reposition the marker only on a machine that's stopped and locked out. Never reach toward a spinning rotor with tape in your hand, at any speed. And before a run, confirm the tape can't come flying off: on fast rotors, secure it with a wrap of heat-resistant tape around the circumference.

## Dirty, oily, hot rotor: what to do

A perfectly clean shaft is something you see in the manual's photographs. On site, what usually waits for you is oil mist, dust, mill scale, or a hot housing. Below are the honest options for each case.

### Oil mist and dust

The marker loses reflectivity gradually: the first readings look fine, and half an hour later the count starts breaking up. Widen the marker, move the sensor closer, and wipe it down at every stop. If the machine is constantly dusty, optics simply loses here: fit an inductive sensor reading off a key or a bolt head instead.

### A hot surface

The adhesive on standard reflective tape is rated for moderate heat, around 60-80°C - check your own tape's packaging for the exact figure. On a hot housing it lifts off and drifts around the circumference, taking your reference point with it. Move the marker to a cooler element on the same shaft: a half-coupling, a free end face, the fan's hub. The ratio to the rotor still has to stay 1:1.

### Mill scale, rust, thick paint buildup

Tape won't stick to any of these. Sand a 20-30 mm strip down to bare metal and degrease it. Where you can't clean down to metal - a coating, a corrosion-protection layer, a warranty condition - use a painted contrast marker on a darkened ring instead, or a non-contact sensor reading off a mechanical feature.

### No direct line of sight with the guard closed

Work this out before the visit, not on site. What usually helps is a half-coupling visible through an inspection window, a free shaft end on the non-drive side, or a hub on the air-intake side. A pulley only works if it's at a 1:1 ratio with the rotor being balanced. If there's genuinely no access at all, agree the solution in advance: an inspection hole in the guard, with the owner's permission, or a method built around extra runs.

> The temptation to fix this with software settings instead of the marker is understandable, but it doesn't work that way. More averaging smooths out noise - it doesn't recover missed pulses. A dropped count is a lost period, and no amount of filtering makes up for it.

## Mounting the laser sensor: distance, angle, fixing

Take the distance to the marker from your sensor's datasheet. For optics reading a reflective tape, that's usually anywhere from a few centimetres to half a metre, and it's better to work in the middle of that range than at its edge. At the limit, any contamination on the marker immediately breaks the count.

Aim the beam nearly perpendicular to the surface, but not exactly: a 5-15° tilt removes the specular glare from the shaft itself, which would otherwise bounce straight back into the receiver and mask the marker. Confirm the beam lands on the marker, and only the marker, throughout the rotor's full travel. The shaft moves axially, grows with heat, and can shift within the bearing during coast-down: something that was on target cold sometimes drifts to the edge of the keyway once the machine is running.

The mount matters more than it looks. Set the magnetic stand on a stationary, rigid surface: the machine bed, the bearing housing's base, the foundation plate. Not on a cover, not on thin guard sheeting, not on a handrail, not on a pipe. A trembling sensor shifts the pulse edge from one revolution to the next, and the phase starts wandering by tens of degrees even at a perfectly stable RPM. In that situation, people go looking for the cause in the rotor, when it's standing on the tripod.

A separate point about light. Direct sunlight into the receiver, pulsed LED lighting, welding nearby, a glare off a wet, oily shaft: any of these adds false triggers. Shade the sensor with your hand or a piece of cardboard and see whether the RPM reading changes. If it does, go looking for stray light, not a fault in the rotor.

- [x] The tripod stands on a stationary, rigid surface, not on a cover or guard
- [x] When the rotor is turned slowly by hand, the beam lands on the marker, and only the marker
- [x] The distance to the marker is in the middle of the rated range, not at its limit
- [x] The cable is secured, and the slack loop neither dangles nor touches any rotating parts
- [x] No direct sunlight or pulsed lighting is shining into the receiver
- [x] There are no shiny bolts, chrome nuts, or a second sticker in the field of view
- [x] The sensor is plugged into its own input (on the Balanset-1A, that's X3), with the accelerometers on X1 and X2

> Don't look into the beam, and don't point it at a colleague's eyes. These sensors carry a low laser class, but it's worth building the right habit from day one.

## Checking before the first run: two minutes that pay for themselves

This check is done before the initial measurement. It costs two minutes and rules out most of the 'the instrument is showing nonsense' stories before they start.

1. **Turn the rotor by hand** — Watch the sensor's indicator light or the pulse in the instrument's window. One pulse per revolution, no more and no fewer. Two pulses mean a second reflection; zero pulses mean the beam is missing the marker, or the distance is outside the sensor's range.
2. **Start the machine and watch the RPM** — Watch the RPM field for a few seconds straight. A stable number with a couple of RPM of scatter is normal. A jump to exactly double means a stray reflection; exactly half means a missed pulse. Drifting by tens of RPM isn't the sensor anymore - that's the drive: a slipping belt, a motor that can't hold the load, or a wandering VFD setpoint.
3. **Cross-check RPM against the drive** — A two-pole induction motor on a 50 Hz supply gives roughly 2900-2970 RPM under load; a four-pole one, roughly 1450-1480. Through a belt drive, calculate it from the pulley diameters: n2 = n1 · D1 / D2. Treat the nameplate as a reference, not as fact: it shows the rated value, not what's actually happening right now.
4. **Cross-check RPM against the spectrum** — The running-speed component's peak should sit at N/60 Hz. For 1478 RPM, that's 24.6 Hz. If the most prominent peak isn't there, you're either looking in the wrong place, or the marker is on the wrong element.
5. **Look at the time waveform alongside the pulses** — On the plot, the vibration sine wave's period should line up with the interval between the marker's pulses. If it lines up, the chain is alive and the sensors and marker are working. If it doesn't, sort that out before you balance anything.
6. **Take three measurements in a row, changing nothing** — RPM should agree within 1-2 RPM, the 1x magnitude within a few percent, and phase within a few degrees. A wider spread means something is unstable: RPM itself, the sensor mounting, the tachometer stand, or you're working close to resonance. You can't balance from data with that much scatter - the influence coefficient will come out essentially random.

> A discrepancy of more than 1-2 RPM between the instrument's display and a standalone tachometer's display is a symptom worth noting on its own. Most often it points to interference in the measurement chain: welding nearby, a VFD with no filter, a power cable run alongside the signal cable. Separate the cables, switch the laptop to battery power, and repeat the measurement.

Sources: [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [ISO 13373-3:2015](https://www.iso.org/standard/40840.html)

## Common problems: symptom, cause, what to do

| What you see | Likely cause | What to do |
| --- | --- | --- |
| RPM reads exactly double what's expected | The sensor sees two reflections per revolution: a second sticker, a shiny key, a bolt head, an oil glare, tape that has wrapped into a closed ring | Turn the rotor by hand and find the second reflection. Darken it with matte paint or tape, and shift the sensor lengthwise if needed |
| RPM reads half, or periodically drops out | Missed pulses: the marker is too narrow for this speed, the sensor is too far away, the marker is dirty, the leading edge has lifted | Move the sensor closer, widen the marker around the circumference, wipe it clean. Check the RPM against the sensor's rated range |
| RPM matches neither the nameplate nor the pulley calculation | The marker is on the wrong element: a rotating cover, the motor pulley at a different ratio, an intermediate shaft | Move the marker to the shaft or hub of the actual rotor being balanced. There's one condition: exactly one revolution of the marker per revolution of the rotor |
| RPM wanders by tens of RPM from one measurement to the next | The belt is slipping, the drive can't hold the load, the VFD setpoint is unstable | Tension the belt, check the drive and the VFD's mode. Neither phase nor the influence coefficient can be calculated from unstable RPM |
| Phase jumps by tens of degrees while RPM is stable | The phase sensor is vibrating on its own: the tripod is on a cover or a non-rigid bracket. Alternatively: operation close to the supports' resonance | Move the tripod to a stationary, rigid surface. If phase keeps wandering, check the machine for resonance during coast-down |
| The count breaks up 10-15 minutes after starting | Oil mist and dust settle on the marker, the tape's edge lifts from heat and centrifugal force | Stop the machine, clean the pad, apply a new marker and press the edge down firmly. On machines that stay dirty all the time, switch to an inductive sensor |
| RPM is stable, but 1x looks bigger than the overall vibration | A measurement error, not physics: 1x is always part of the overall level. Unstable RPM, a stray reflection, or an overloaded input are usually to blame | Sort out the pulse count and the signal level first. There's no point balancing before that's fixed |

> The general rule: get stable RPM first, then stable phase, and only then take the initial measurement. Each step depends on the one before it, and doing this out of order is guaranteed to cost you extra runs.

## The marker sets the phase zero, not the correction-angle zero

This is where confusion happens most often, and it's an expensive one. The reflective marker sets the reference point for measuring phase. The trial weight's location sets the zero point for measuring the correction weight's mounting angle. These are two different zero points, and they don't have to coincide.

The practical conclusion follows from that: if the instrument shows F1 = 47°, that doesn't mean the heavy point sits 47° from the marker. The phase reading depends on where the accelerometer is mounted and which way it points, which edge the tachometer triggers on, how stiff the supports are, and how close you are to resonance. The exact same extra mass on two similar machines will give you different degree readings.

What actually matters about the marker is consistency. As long as the reference point doesn't move, every phase reading stays comparable, and the 'before and after the trial weight' vector difference means something. Shift the marker between the initial run and the trial run, and every phase reading shifts by the same fixed amount, and the influence coefficient comes out wrong. Spotting this from the numbers alone is practically impossible - they look perfectly normal.

That's why the marker goes on once, before the initial measurement, and stays untouched until the job is finished. If it does come loose, restore it against the line you traced along the leading edge with a marker pen. If you can't restore it, and the correction hasn't been calculated yet, retake the initial measurement and the trial runs from scratch.

And one trick that saves you on the next visit. Mount the trial weight at the same angular position as the marker. Then the angle-reference zero coincides with the marker, it's always visible, and saved influence coefficients can be reused a year later. We cover this in more detail in the articles on angle referencing for weight mounting, and on trim balancing.

> Don't compare phase readings taken on different visits if the marker or the sensors were positioned differently. Strictly speaking, the degree values can only be compared under identical setup: the same accelerometer locations, the same measurement direction, the same angular position for the marker, the same RPM.

## Alternatives to the laser sensor: encoder, key, gear tooth

Optics reading a reflective marker is the fastest way to get a reference pulse, but not the only one. When tape won't hold, or the shaft isn't visible, other options work.

### Inductive or eddy-current sensor on a key

Responds to a steel protrusion or notch: a key, a bolt head, a set screw. It doesn't care about oil, dust, light, or moderate heat, so on dirty and hot machines it's often the only option that actually works. The cost: it needs a gap of about 1-3 mm, a rigid bracket, and exactly one mechanical feature per revolution.

### Sensor on a gear tooth

It's tempting to point the sensor at a gear's teeth, but a gear with z teeth gives you z pulses per revolution, and balancing needs exactly one. A signal like that only works if you can pick out a single distinguishing feature: a lengthened tooth, a missing tooth, a target screw fitted into the gear. Otherwise the instrument will read RPM as z times too high.

### Encoder

A multi-pulse sensor on the shaft end. You need it when order analysis under varying RPM, instantaneous rotation speed, or torsional vibration actually matter. For routine balancing its resolution is overkill, and fitting it needs a coupling and access to the shaft end. It's usually found on test stands, not out in the field.

### The machine's own built-in reference sensor

On turbines and compressors with plain bearings, a reference sensor reading a groove machined into the shaft is often already installed as part of the monitoring system. If it has an accessible signal output, it makes more sense to use that ready-made reference point than to stick tape inside the housing.

### A painted contrast marker

Some optical sensors work on contrast rather than retroreflection: a white painted stripe over a darkened ring. The signal is weaker, and the working distance shorter, but it holds up where tape simply won't stick.

### Stroboscope

Lets your eyes see the marker as if it were standing still, and lets you read off RPM. But it never sends a pulse to the instrument, so there's no 1x, no phase, and no weight calculation coming from it. It's a tool for inspection, not measurement.

> On the Balanset-1A, input X3 expects a pulse from its own laser phase sensor, so agree on any substitution in advance: the level, polarity, pulse duration, and power supply all need to match. In the Balanset-1A OEM version, which gets built into machine tools and test stands, the phase sensor usually becomes a permanent part of the structure: a bracket, a marker on the spindle or drive shaft, a fixed distance. Kits like that are put together for a specific piece of machinery.

## If it's simpler to have us come out with the instrument

The phase sensor is the one part of the system where mistakes don't look like mistakes. The instrument won't tell you 'the marker got dirty' - it will calmly calculate a weight from the wrong phase, and you'll only see the result on the verification run. On an easily accessible fan, that's one extra run and half an hour lost. On a unit that only gets shut down once a quarter, up at height, or on a hot line, there usually isn't a second attempt.

AXILINE's engineers design and manufacture the Balanset instruments and use them to balance rotors on site themselves. We arrive at the machine, mount the accelerometers on the bearing housings, and set up the laser sensor against a reflective marker. First we check the pulse count and RPM stability, then we check whether the vibration is actually coming from imbalance rather than shaft misalignment, loose fasteners, or resonance. Only after that do we balance the rotor in its own bearings: in one or two planes, with weights or by drilling, by angle or by fixed positions. We hand over the result as a report with the initial and residual running-speed component, and a note of where, and how much, mass was fitted.

If you're balancing it yourself, consulting support on the instrument comes bundled with the Balanset-1A, including exactly this kind of question: where to find a reference point on your machine, what to use instead of tape on a hot shaft, how to check the count. As for the result: a typical order of magnitude for a fan with a dominant 1x peak is a drop from roughly 12 mm/s to roughly 1.5-2 mm/s. Nobody promises that figure in advance, because the outcome depends on the condition of the supports and fasteners, and on whether imbalance is the only thing driving the vibration. Judge the machine's overall condition from overall vibration in mm/s RMS against the applicable part of ISO 20816, and the residual-unbalance tolerance from the grade-G tables in the applicable part of ISO 21940.

Sources: [Balanset-1A manufacturer specification](https://vibromera.eu/product/balanset-1/) · [Balanset-1A operation manual](https://vibromera.eu/balanset-1a-operation-manual/) · [ISO 20816-1:2016](https://www.iso.org/standard/63180.html) · [ISO 21940-11:2016](https://www.iso.org/standard/54074.html)

## Frequently asked questions

**Can a rotor be balanced without a phase sensor?**

Practically no, not for the normal way of doing this. Without a reference pulse, the instrument only shows overall vibration: there's nothing to extract the running-speed component from, nothing to measure phase against, and so nothing to subtract when calculating the influence coefficient. Methods based on amplitude alone do exist, where the trial weight is moved sequentially between marked positions, but every position is a separate run with its own run-up and coast-down. On live production equipment, that kind of saving ends up costing more than the sensor.

**Where should the reflective marker go, and what size should it be?**

On the shaft, the impeller hub, a half-coupling, or a spindle end face: any element that makes exactly one revolution for every one revolution of the rotor being balanced. Degrease the surface, and clean it down further if needed. Working guide for size: 10-20 mm around the circumference and 8-15 mm lengthwise. The higher the peripheral speed, the wider the marker should be: on a Ø300 mm hub at 3000 RPM, a 10 mm strip passes the beam in 0.2 ms.

**The instrument shows double the expected RPM. What's wrong?**

The sensor is seeing two reflections per revolution. Turn the rotor by hand and find the second source: an old sticker, a shiny key, a bolt head, a chrome nut, an oil glare, or tape that has wrapped into a closed ring on a thin shaft. Darken that second reflection with matte paint or tape, or shift the sensor lengthwise. RPM reading at half the expected value means the opposite: the instrument is missing pulses - the marker is narrow or dirty, or the sensor is too far away.

**Does the marker set the correction weight's mounting angle?**

No. The marker sets the reference point for measuring phase. The weight's mounting angle is measured from wherever the trial weight was - that's the 0° point. The F1 phase you see on screen isn't the heavy point's angle: it depends on where the accelerometer is mounted and which way it points, which edge the tachometer triggers on, how stiff the supports are, and how close you are to resonance. A useful trick: mount the trial weight at the same angular position as the marker, so both zero points coincide, and saved influence coefficients will still be useful on the next visit.

**The rotor is hot and covered in oil, and the tape won't hold. What now?**

First try moving the reference point to a cooler, cleaner element on the same shaft: a half-coupling, a free end face, a hub. The ratio to the rotor still has to be 1:1 - a pulley at a different ratio won't do. If there's genuinely no clean spot anywhere, switch to a non-contact inductive sensor reading off a key, a bolt head, or a set screw: it doesn't care about oil, dust, or heat. The adhesive on standard reflective tape holds up to moderate heat, around 60-80°C - check your tape's packaging for the exact figure.

**Phase jumps by tens of degrees even though RPM is stable. Where should I look?**

Start by checking the phase sensor itself. A magnetic stand on a cover, a guard, or thin sheeting vibrates along with the machine, the pulse edge shifts from one revolution to the next, and phase wanders as a result. Move the tripod onto the machine bed, the bearing housing's base, or the foundation plate. If phase keeps drifting after that, check the machine for resonance: within a resonance band, the running-speed component's phase swings by almost 180° for a small change in RPM, and you won't get stable balancing results there.