Speed sensor selection and installation considerations for trouble-free operation

Selecting the right sensor for overspeed protection, speed indication, and speed control starts with understanding the measurements, understanding the different types of sensors, and understanding application details that can influence measurement integrity.


Consider the consequences of a “missed” overspeed event on a large steam turbine-generator. In less than a second, the machine can accelerate to speeds that liberate blades from the rotor, puncturing the turbine case and creating enormous damage to not just the machine, but surrounding structures, equipment, and personnel. Figure 1 is a sobering reminder of the consequences and requires little elaboration1,2,3


Consider also the implications of faulty speed control applications. One example is the “zero speed” measurement and this is more fully detailed in our eBook on TSI measurements. Suffice to say here that it must accurately measure very slow speeds4 so that when a rotor has slowed sufficiently, the turning gear can be engaged to prevent a rotor bow from occurring. Engagement of a turning gear at the wrong speed can result in severe damage to the rotor and turning gear apparatus. Likewise, failure to disengage the turning gear as a turbine begins ramping up can also result in damage. Indeed, there are both economic implications and safety implications. Whether done manually or automatically, a system measuring speed for turning gear engagement/disengagement is serious business.

Governor control is obviously also important with the ability to control speed very precisely in a variety of applications, not the least of which is power generation when synchronizing the rotative speed of a generator with the grid. Another example is the proper ramp rate of a turbine as it starts up so that thermal expansion can occur without incurring rubs. This topic is also addressed in our eBook on TSI measurements.

overspeed incident

Figure 1: Aftermath of February 2011 overspeed failure on Unit #4, one of six 600MW turbine-generator units at the Duvha Power Station in South Africa1,2,3. While the unit is designed for 60Hz (3600 rpm) operation, the last recorded speed before destruction ensued was 4250 rpm (18% overspeed). Ironically, the failure occurred during testing of the unit's overspeed trip system.


In large hydro units, so-called “creep” measurement5 is highly critical as it must determine whether a rotor has come to a complete standstill, allowing personnel to safely service the massive machine without being injured or crushed. Creep occurs when water leakage trickles through a turbine even with fully closed valves or gates, and the rotor braking system is not strong enough to maintain a total shaft standstill. This leakage slowly rotates the machine, acting like an unintended, hydraulic turning gear. The speeds are so low (typically 3 revolutions per hour – a mere 0.05 rpm) that a conventional speed measurement is inadequate. For example, use of the phase reference signal would result in only a single pulse every 20 minutes – an unacceptably slow update rate when human safety is at stake. Various approaches for this measurement exist. Some use a 360-tooth gear on the shaft and a conventional proximity probe, providing one pulse for every degree of revolution and an update rate of about 10 seconds. Others use an air gap sensor. Still others use mechanical apparatus such as a small wheel that contacts the shaft periphery and rotates faster than the turbine shaft due to the ratio in diameters6. Other approaches use a brush that contacts the shaft. As the shaft rotates, the brush tilts and toggles a microswitch, signalling that creep is occurring. Regardless of the mechanism used, the integrity of the creep measurement is extremely important as the safety of maintenance personnel is at stake.

Reverse Rotation:

Another important measurement is that of reverse rotation, which can occur when the process fluid flows backwards through a machine, causing it to rotate in the wrong direction. This can occur on compressors, for example, turning the compressor into a driver and the prime mover into the driven machine. For machines with dry gas seals, this can be especially destructive as they are designed to incur rotation in only one direction and can only sustain a certain number of reverse rotations before damage and corresponding levels of unacceptable seal leakage occur.


Figure 2: One method of making a reverse rotation measurement is by means of two proximity probes. In the top figure, the gray probe observes the once-per-turn shaft discontinuity before the green probe, indicating clockwise rotation. In the bottom figure, rotation is occurring in the counterclockwise direction. and the green probe thus observes discontinuity before the gray probe. Other schemes are also possibly, such as through the use of hall-effect sensors with two embedded magnets that allow direction of rotation to be ascertained from a single sensor.


Lastly, we have speed indication which is – as the name implies – the indication of speed rather than using it for protective or control purposes. Speed indication, although important, does not have the safety or economic consequences associated with these other speed measurements. As such, the redundancy of sensors and measuring circuits, and the testing intervals, are not nearly so vital.

For all of these reasons, understanding the sensors used to measure speed and the application considerations that favor one type over another are important. Also important is proper attention to details in the overall overspeed system and field wiring that can – if not observed – result in measurement vulnerability even when the proper sensor has been selected.

Speed Sensor Types

There are three basic types of speed sensors used for industrial machinery measurements: eddy-current proximity, magnetic, and Hall-effect. Although there are others, they comprise only a very tiny portion of the installed base for speed measurements on turbomachinery and are thus not discussed here.

Eddy Current Proximity Sensors

Users of vibration monitoring systems such as the VM600 will be familiar with these transducers because they are used extensively for radial vibration, axial position, and phase reference measurements.

Proximity probes are a good choice when precise phase measurements must be made, such as crank angle on reciprocating compressors (where torsional vibration is often present and non-uniform speed throughout the stoke) and on any machine where diagnostics will be performed using phase. For reciprocating compressors, a precision measurement wheel is often used as detailed in API 670 Annex J.

The small size of proximity probes (down to 5mm tip diameter and M6 or 1/4-28 UNF threads) allows them to be mounted in constrained spaces, which can be important on machines with small shaft diameters such as expander-compressors and integrally geared compressors with multiple pinions.

You can read more about the use of proximity probes for speed measurement in our blog article Proximity Measurement Systems as Speed Sensors.

Magnetic Sensors

These are variously known as speed pickups, magnetic sensors, magnetic pickups (MPUs), passive magnetic sensors, and variable reluctance (VR) sensors. They are by far the most common type of speed sensor for governor control and overspeed measurement. However, they are generally not suitable for low-speed machinery where rotational speeds lower than 250 rpm are required (unless a multi-tooth speed wheel is used for the observed surface). Unlike Hall-Effect and Eddy-Current proximity sensors, they are not capable of true zero-speed measurements. Also, unlike Hall-Effect and Eddy-Current proximity sensors, they require a target that is ferrous.

MPUs employ a coil encircling a magnet and measure a change in magnetic flux produced by discontinuities in the target surface (such as gear teeth passing under the sensor). They thus require a ferrous target. The output from an MPU is a sinusoidal signal and the amplitude depends on the distance (gap), the size of the target, and the speed. This can be undesirable at both very high speeds and very low speeds. At high speeds, the voltages can be large – exceeding 60Vac – and well beyond typical instrumentation voltages of 24V or less. In contrast, at low speeds, the voltages are generally too small to use without additional amplification. This generally limits passive MPUs to measurements above 250 rpm.

The typical size used in many turbines is 5/8-18 UNF-2A or M16, but smaller sizes are available down to ¼-28 UNF-2A or M12-1.25.

Because the sensors are self-generating and require no external power, many turbine manufacturers standardized on them and did not provide their corresponding control/monitoring/protection systems with the ability to power a speed sensor. This means that other sensor types cannot easily be substituted without upgrading the control/monitoring/protection system. Consequently, most users replace their MPUs with a like-with-like product rather than switching to a Hall-Effect or Eddy-current sensor.

Popular manufacturers of MPUs include Dynalco (Barksdale), Magnetic Sensors Corporation, HarcoSemco, and Motion Sensors Inc. However, there are many others and this is only a partial list.

Active (i.e., externally powered) MPUs are also available that can provide a digital output (i.e., rectangular pulse train) instead of a sinusoidal output that varies in amplitude instead of only in frequency. Although active MPUs can better address low-speed measurements (down to about 2Hz), they are still not capable of making true zero-speed measurements.

You can read more about the working principles of MPUs in a white paper7 published by HarcoSemco, available for download here.

Hall-Effect Sensors:


Like an MPU, Hall-Effect sensors also use an embedded magnet and thus require a ferrous target. Although both types of sensors measure magnetic flux, the Hall-Effect sensor is only sensitive to the amount of magnetic flux – not its rate of change. Consequently, the amplitude of the output signal does not change with speed – only the frequency changes. However, because Hall-Effect sensors – unlike proximity probes and MPUs – have embedded electronics instead of relatively simple components like a magnet and/ or coil of wire, they are limited to environments with temperatures below about 150° C. Although the technology has advantages and is quite robust, not nearly as many Hall-Effect sensors are used on industrial turbomachinery as are MPUs or eddy-current proximity probes.

You can read more about the working principles of Hall-Effect sensors in a white paper8 published by Honeywell, available for download here.

Sensing Surfaces

While a once-per-turn discontinuity is used for phase reference measurements, and is suitable for most indication-only speed applications, it is emphatically not suitable for overspeed measurements. If one considers that API 670 requires the overspeed protection system to act within 40ms of sensing an overspeed event, a once-per-turn speed signal takes far too long to update. Consider a 3000 rpm (50 Hz) machine. A single shaft revolution requires 20 ms. If we were to require even two successive measurements to sense an increase in speed, this itself would consume nearly 40 ms. For slower-speed machines, the situation becomes even more unfavorable. For this reason, it is common to use a toothed wheel so that speed changes can be detected as quickly as possible – generally within just a fraction of a full rotation of the shaft.

In some cases, a specially manufactured gear is used, designed and optimized specifically for making speed measurements. In other cases, an existing gear is used. The profile of the gear is important, particularly for MPUs. API 670 provides guidance on this in Annex J9 where minimum, nominal, and maximum dimensions are given for both precision speed sensing wheels (Figure 3) and non-precision (i.e. gear) speed sensing wheels (Figure 4). Due to copyright limitations, the figures themselves are reproduced here by permission, but a copy of the API 670 standard must be purchased10 to access the actual minimum, nominal, and maximum recommendations for these dimensions.

Figure 3: Relevant dimensions for precision speed sensing surfaces.


Figure 411: Relevant dimensions for non-precision (i.e., gear) speed sensing surfaces.

Common Pitfalls


Figure 5: If a rotor grows by an amount Δ, care must be taken to ensure it does not result in the target surface moving outside of the transducer's observed field.


  1. Excessive rotor thermal growth at sensing surface - The location of the speed measurement surface must account for thermal growth of the shaft, which can be particularly pronounced on large machines such as steam turbine-generators. If the shaft grows or shrinks considerably at the measurement location, the sensing surface can move outside the transducer’s observable field, resulting in an erroneous measurement. See Figure 5.
  2. Excessive field wiring length - When field wiring lengths exceed several hundred meters, the distributed impedance can become appreciable. The wire then acts like a low-pass filter, attenuating high frequency signals. Care must be taken to understand and account for these effects or to use a sensor that is less prone to such issues, such as vibro-meter’s TQ family of proximity probes and selecting the dynamic current output option rather than the dynamic voltage output option.
  3. Improper probe gap - Improperly gapped probes can either rub and be destroyed altogether during high vibration of the observed surface, or can give erroneous readings. MPUs give a decreased amplitude with increasing gap. One of the advantages of an active probe is that it is able to provide more robust OK checks than an unbiased sensor such as an MPU. A proximity probe will normally be biased at approximately -10V when energized with conventional -24Vdc power. It is very common for the toothed wheel to exhibit larger eccentricity (radial vibration amplitudes) during an overspeed event and wipe the tip of the sensor. Even if redundant sensors are used (which is very common for overspeed measurements), all sensors may be wiped within a single shaft revolution if not gapped properly, thus rendering the overspeed system inoperable. This is one of the reasons that an overspeed system treats a NOT OK sensor as a vote to trip. It is also why separate sensors and even separate sensing surfaces are used for governor control versus overspeed protection.
  4. Improper trigger settings - For proximity probes and MPUs, the signal output is sinusoidal and is thus a composite of the changing gap due to vibration and the changing gap due to discontinuities in the target (e.g., gear teeth). There are many factors that can combine to give a different signal at operating speeds than at slow-roll speeds and it is important to set triggering levels that reflect operating speeds – not merely slow-roll speeds. It is strongly recommended that an oscilloscope be attached to the speed sensor’s output in order to observe the actual waveform at operating speeds and to set trigger thresholds accordingly. Figure 6 illustrates what can happen when triggering is set by observing only a slow-roll waveform, resulting in erroneous readings (undercounting) at operating speeds. The reason why a signal at slow-roll speed may have a larger amplitude than at operating speed reflects numerous factors including transducer slew rate, frequency response, and the effect of vibration amplitude and phase at the measurement location.


Figure 6: An example of an incorrect triggering level VT from a proximity probe. While it works acceptably at slow-roll speeds (top), it does not work acceptably at operating speeds (bottom) where only one of the four pulsed is recognised, resulting in an erroneous speed reading. Undercounting of actual gear teeth will result in a missed trip from an overspeed system. Conversely, overcounting of gear teeth will result in a false trip.

Improper surface profile

As already noted, and as conveyed in Figures 3 and 4 in conjunction with API 670, the dimensions of the observed surface profile must be carefully adhered to. Figure 7 is an example of the output from an MPU where these considerations were not observed, resulting in a so-called “multiple zero crossing” and leading to an erroneous speed indication.


Figure 7: An example of an MPU observing a proper surface (top) versus an improper surface (bottom). An improper surface can lead to multiple positive-going zero crossings and thus erroneous readings. The self generating nature of an MPU produces an AC waveform that is centered on 0V ( i.e., it has no bias voltage).


These are by no means the only potential pitfalls when making speed measurements – they are merely five of the most commonly encountered. Consult your vibro-meter service professional for additional guidance and installation assistance.


While vibration, axial position, and other measurements are used for machinery protection purposes, no measurement exceeds that of speed in its importance and its potential for catastrophic machinery failure if not implemented properly. While transducer selection is important, installation is equally important along with attention to the observed sensing surface. Even the most capable overspeed detection instrument coupled with the correct transducers will be rendered ineffective unless the installation pitfalls outlined here – as well as others that can be identified by your vibro-meter service professional – are recognized and avoided.


1Gabara, N., “Repairs to Eskom's Duvha station to take time”, South African Government News Agency https://www.sanews.gov.za/south-africa/repairs-eskoms-duvha-station-take-time (Feb 10, 2011)

2The entire collection of images is available at http://www.tathasta.com/2018/03/duvha-south-africa-turbine-overspeed.html

3Straton, A., “Duvha Power Station – an Exercise in Incredulity”, www.MyPE.co.za, https://mype.co.za/new/duvha-power-station-an-exercise-in-incredulity/3439/2011/03/ (Mar 26, 2011)

4Turning gears typically run at 3 rpm or less.

5Not to be confused with blade creep on gas and steam turbines which is the gradual distortion (elongation) of blades due to the large centrifugal forces imparted over time from high-speed blade rotation. Such creep is normally described in terms of microinches per operating hour.

6For example, a 1m turbine shaft diameter contacting a 5 cm diameter wheel will cause the wheel to rotate 20 times faster than the turbine shaft. For a turbine creeping at 3 revolutions per hour, this translates to a measurement wheel rotating at 1 rpm.

7Croce, R.A., Giterman, I., “Development of the Electrical and Magnetic Model of Variable Reluctance Speed Sensors”, white paper, Nov 2016, Harco Laboratories, Brandord, CT.

8Hall Effect Sensing and Application”, white paper, Oct 1998, Honeywell MicroSwitch Sensing and Control, Freeport, IL.

9API Std 670 – Annex J, “Machinery Protection Systems – Electronic Overspeed Detection System Considerations,” 5th Edition, Nov 2014, American Petroleum Institute, Washington, DC.

10API standards, technical reports, recommended practices, and other publications can be purchased at techstreet.com, IHSMarkit.com, and other authorized distributors globally (https://www.api.org/products-and-services/standards/purchase#tab-authorized-standards-distributors).

11These figures are used with permission of American Petroleum Institute and appear in Annex J of API Standard 670, Machinery Protection Systems, 5th edition, (Nov 2014).

Back to blog list

Learn more