Noncontact Rotational Speed Measurement with Near-Field Microwave of Open-Ended Waveguide

Abstract

Rotational speed measurement is important for many applications. Here, a noncontact rotational speed test method based on the detection of the periodically perturbed near-field microwave of an open-ended waveguide is proposed. Both simulations and experiments were conducted to verify the near-field microwave rotational speed sensor. The constructed rotation speed sensing system was composed of a standard open-ended WR-42 waveguide (in our measurements, a waveguide-to-coaxial adapter was used to represent an open-ended waveguide) working at ~18 GHz, a radio frequency (RF) circulator, a signal generator, a, RF detector and an oscilloscope. A rotating fan to be measured was placed close to the waveguide’s mouth and, thus, the waveguide’s reflection coefficient was periodically modulated by the rotating fan blades. Then, the RF detector converted this varying reflection coefficient into a direct current (DC) voltage, namely, a periodical waveform. Finally, the rotational speed of the fan could be extracted from this waveform. Measurements using both the proposed near-field microwave method and conventional optical transmission/reflection methods were conducted for verification. The effect of the rotating fan’s location relative to the waveguide’s mouth was also studied. The results show the following: 1. The proposed method works well with a rotational speed of up to ~5000 RPM (rounds per minute), and an accuracy of 1.7% can be achieved. 2. Metallic or non-metallic fan blades are all suitable for this method. Compared with the existing radar method, the proposed method may be advantageous for rotation detection in a constrained space.

1. Introduction

The rotational speed measurement of rotating parts is important for applications with, for example, generators, electric motors, machine tool spindles, engines and turbo pumps [1,2,3,4]. There are different kinds of rotational speed measurement methods, including electrostatic sensing [5], using a capacitive angular speed sensor [6], using a wireless passive LC sensor [7], the magnetic induction-based method [8] and using an IR-UWB radar [9], and they can be classified into several subgroups like optical, electrical, magnetic and electromagnetic techniques. Among them, electromagnetic methods possess some advantages over other methods such as noncontact ability (suitable for on-line monitoring) and penetrability (electromagnetic waves at RF/microwave bands can penetrate non-metallic material, making it possible to detect rotating parts behind/inside opaque obstruction which cannot be detected by optical methods). For example, inductor–capacitor sensors, consisting of two coils attached to the stator and the rotor, were presented in [1,2] for rotational speed measurement. Similarly, different kinds of microwave sensors are reported in [10,11,12,13,14,15,16,17,18,19] for the detection of angular displacement. In [10], a novel two-way rotation sensor is proposed deploying substrate integrated waveguide (SIW), and the rotation angle can be extracted from resonance frequency shift and notch dip magnitude variation. In [11], a novel rotation sensor based on the concept of the cross-polarized excitation of split ring resonators (SRRs) is presented and the rotation angle is sensed by measuring the phase difference of the reflection coefficients of the two ports of a slot-line loaded with a rotatable SRR. In [12], a fixed-frequency angular displacement sensor using a dielectric-loaded metal strip resonator is reported and the rotation angle is obtained from |S₂₁| at a single frequency. In [13], a microwave planar sensor, consisting of a movable modified CSRR (the rotor) and a microstrip line with a circular defect in the ground plane (the stator) is proposed to measure the angle of rotation in terms of the change in the relative phase of the reflection coefficients. In [14], based on a CSRR driven by a conductor-backed coplanar waveguide, a novel sensor for detecting and measuring angular rotation and proximity, intended for rapid prototyping machines, is presented. In [15], using the SRR and CSRR as the rotor and a modified convex-shaped open CSRR as the stator, an active feedback microwave sensor with a wide dynamic range and high sensitivity for detecting angular displacement is presented. In [17], an ultra-wideband (UWB) bandpass-filtering microstrip circuit with an embedded in-band transmission zero for co-integrated angular displacement RF sensing is reported. In [16,18,19], some kinds of SRRs or CSRRs are used for rotation sensing. Most of the sensors mentioned above can be classified as near-field methods and they usually require a specific rotor for the purpose of angular displacement detection. This may limit their applications when it is undesirable/impossible to attach some elements onto rotating parts. Besides these near-field methods, far-field methods for rotation sensing have also been studied. In [9], a UWB radar was used for rotational speed tests, with the obvious advantage of remote detection (0.5 m) compared with near-field microwave methods. In [20], a contactless rotation speed measurement system based on a millimeter-wave radar was reported and the proposed system calculated the cycle period of the rotator by measuring the cycle of the relative linear distance between the radar and the scattering points on the blades of the rotator. In [21], a method for measuring the rotational speed of industrial wind turbine blades using a 24 GHz Doppler radar was proposed. In [22], a practical millimeter-wave radar-based rotation speed sensing system was proposed and was able to separate the target signal reflected by the rotating object from the mixed reflection signals, extracting high-quality rotation-related features. However, sensing with a radar usually requires the specific consideration of signal processing and a large operation distance. The latter requirement may inhibit its usage in cases with a limited operation space, like inside a gearbox.

Here, we propose a near-field microwave method pertaining to rotational speed evaluation. The method has merits in the following aspects. First, there is no need to attach elements to the rotor. Second, its signal processing is relatively simple (only a simple filtering process is required). Third, its working distance can be much shorter than its counterparts based on far-field methods like using a radar. We used a microwave radiator (an open-ended waveguide) to interact with the rotating fan in the test, which was excited with a continuous sinusoidal signal. When the fan blade was close to the open-ended waveguide’s mouth, it perturbated the waveguide’s near-field and changed its reflection coefficient. If the blade moved away from the waveguide’s mouth, this perturbation disappeared. Thus, when a rotating fan was close to the waveguide’s mouth, the reflection coefficient of the waveguide changed periodically, and its periodicity was inversely proportional to the rotational speed. So, one could extract the rotational speed from the periodicity of the waveform of the reflection coefficient.

In Section 2, we present the principle and simulation results. In Section 3, we give a description of our test setups and the experimental results and discussions. Finally, some conclusions are given in Section 4.

2. Principle and Simulations

2.1. Principle of the Proposed Measurement Method

The basic idea for the proposed rotational speed measurement method is shown schematically in Figure 1. A microwave spot generated by the near-field of a microwave sensor (here, an open-ended waveguide) is represented by a red spot in Figure 1. Namely, at the plane of the fan blade, the energy of the near-field microwave had a specific distribution and could be focused to some degree in the area represented by the red spot. When the fan blade experienced clockwise rotation, as shown in Figure 1a–f, the microwave spot intersected with the fan blade periodically. Since the blade’s material had different electromagnetic properties (like permittivity and electrical conductivity) in air, this meant that the material property of the near-field space of the microwave sensor changed in a periodic way. Finally, the impedance matching condition of the microwave sensor varied with blade position, which could be measured by monitoring the time-dependent S-parameters of the microwave sensor. In other words, one could extract the fan’s rotational speed from the periodicity of the sensor’s S-parameters. In this work, an open-ended waveguide was used as a rotational speed sensor and the magnitude of its S₁₁ was used for direct measurement. For the example shown in Figure 1, since the blade number was three, the sensor’s near-field interacted with the fan blade three times in one round. Thus, one could expect that the S-parameters would change three times for each round. Suppose that the rotation period of the fan was T (its unit is seconds); the rotational speed V_r of the fan in RPM (rounds per minute) could be calculated as follows:

$$V_r=60/T$$

(1)

2.2. Simulations

A number of full-wave electromagnetic simulations were conducted to verify the idea described in Section 2.1. A schematic cross-section view of the adopted electromagnetic model is shown in Figure 2a. An open-ended waveguide was used as a sensing device. In this work, we mainly present the results obtained from a standard WR-42 waveguide. However, standard waveguides with different dimensions were also investigated and similar results were obtained but are not presented here. A rotating fan was placed close to the waveguide flange (thickness 4 mm and edge length 22.4 mm) to put the fan blade in the near-field zone of the open-ended waveguide. A perspective view of the electromagnetic model is shown in Figure 2b. The direction of the adopted Cartesian coordinate system was also defined. We defined the ‘distance’ between the fan and the sensor as the distance along the z axis. The alignment between the sensor and the fan was defined as the relative position in the xoy plane. Definitions of various sizes/dimensions that were used later are shown in Figure 2c,d. For simplicity, the fan had three rectangular blades and they were equally distributed. Namely, the angle between the adjacent blade was 120°. The symbol L₂ is used to represent the distance between the sensor and the fan. The quantities Δ_x and Δ_y are used to describe the alignment between the sensor and the fan. When Δ_x = 0 and Δ_y = 0, this meant that the sensor was aligned with the fan’s center. Typical values of sizes/dimensions are listed in

Table 1. Sizes/dimensions of electromagnetic models. Unit is mm.

Symbol	Values	Symbol	Values
L1	11.00	w	2.00
L2	5.00	r1	15.00
L3	6.00	r2	3.00
L4	10.00	Δx	0.00
h	1.00	Δy	0.00

. If not mentioned specifically, these values were used in our simulations.

In Figure 3, simulated electric near-field distributions of an open-ended waveguide with/without a fan are shown. It should be noted that, different from the fan shown in Figure 1 and Figure 2, a six-blade fan was used here. Also, different from Table 1, in these simulations, Δ_x was 10 mm and Δ_y was 5 mm. L₂ was set as 1 mm. The observation frequency was 22 GHz. The fan’s material was alumina with a relative dielectric constant ε_r equal to 9.9. As shown in Figure 3a–c, the six-blade fan rotated from 0° to 40° in a counter-clockwise direction with a step of 20°. It can be seen that when the rotation angle was 0°, one of the six blades was positioned close to the center of the microwave spot. For 20° and 40°, the overlapping between the blade and the microwave spot was reduced. So, one could expect that, compared with the 20° and 40° cases, the fan would have a stronger influence on the reflection coefficient of the open-ended waveguide when it was at the 0° position. As a reference, the near-field distribution of the open-ended waveguide without a fan is shown in Figure 3d. Thus, the simulation results shown in Figure 3 prove the concept presented in Figure 1. For the rotating fan, as the blade’s position was time-dependent and changed in a periodic way, it could be inferred that the open-ended waveguide’s S₁₁ was also time-dependent and varied periodically. It was exactly this periodicity from which one could extract the fan’s rotation speed or angular velocity.

In Figure 4, the simulated rotation angle-dependent S₁₁ is shown. As shown in the inset, we used a three-blade aluminum fan in these simulations. Also, different from Table 1, in these simulations, Δ_x was 8 mm. For comparison, we used three L₂, namely, 2, 5 and 8 mm, as shown in Figure 4. Other parameters were the same as those in Table 1. The rotation angle θ_r was swept from 0° to 360° with a step of 15°. The excitation frequency of the open-ended waveguide was swept from 18 to 26 GHz, which is the recommended working frequency band of a standard WR-42 waveguide. For a given parameter combination of L₂ and θ_r, a frequency-dependent S₁₁ was obtained. For a given L₂, there were 25 S₁₁ curves (corresponding to θ_r ranging from 0° to 360° with a step of 15°). From these curves, we selected the optimum observation frequency to maximize the amplitude variation of S₁₁. A larger amplitude variation of S₁₁ meant a higher sensitivity of the sensor. Finally, the optimum observation frequency for 2, 5 and 8 mm distances for L₂ was 18.81, 18.44 and 23.08 GHz, respectively. After this post-processing of the data obtained from the simulations, we obtained the three periodical curves shown in Figure 4. It can be seen that in the swept angle range, three periods are presented. This agrees with the number of blades of the used fan. For the three distances, they presented similar amplitude variations, namely, ~2 to 3 dB. For the three observation frequencies (~20 GHz), the corresponding wavelength was ~15 mm. So, all of the considered three distances for L₂ were shorter than one wavelength. Since the amplitude variation of S₁₁ was similar for the considered three L₂, this indicated they had similar sensitivity. By transferring the rotation angle into time (this transformation depended on the fan’s rotation speed), one could obtain periodical time-dependent waveforms (Figure 4), similar to the curves experimentally observed, as shown later.

We conducted a group of simulations to demonstrate the excitation frequency’s effect on the sensor’s sensitivity, as shown in Figure 5. The simulation model and settings were the same as in Figure 4. In Figure 5a, L₂ is 8 mm and three excitation frequencies are compared, namely, 18, 22 and 26 GHz. From the rotation angle-dependent |S₁₁| waveform shown in Figure 5a, one can see that the amplitude variation of S₁₁ depended on the excitation frequency. For example, the variation of |S₁₁| was highest (3.14 dB) at 22 GHz and lowest (0.59 dB) at 18 GHz. To show this excitation frequency effect systematically, we post-processed the simulated results on the whole frequency band, namely, from 18 to 26 GHz, and the obtained results are shown in Figure 5b. It can be seen that for a given L₂, the excitation frequency influenced the amplitude variation of S₁₁. For example, when L₂ was 2 mm, the highest variation occurred at ~18 GHz (~4.5 dB), while the lowest variation occurred at ~21 GHz (~1.5 dB). Thus, from a systematic view, the selection of working frequency should be carefully considered.

In Figure 6, the effect of the distance between the sensor and the fan L₂ is shown. We also used a three-blade aluminum fan in these simulations. Also, different from Table 1, in these simulations, Δ_x was 10 mm. For comparison, we swept L₂ from 2 to 15 mm. Other parameters were the same as those in Table 1. Similar to Figure 5, one can see that the amplitude variation of S₁₁ changed with the excitation frequency of the microwave sensor. However, the focus here was the dependence of ΔS₁₁ on the distance parameter L₂. Generally speaking, ΔS₁₁ decreased with increasing distance L₂. For example, when L₂ was 2 mm, the maximum value of ΔS₁₁ was ~5.5 dB at ~23 GHz, while it was ~2 dB at ~25 GHz when L₂ was 15 mm. On the other hand, the minimum value of ΔS₁₁ (defined as the difference between maximum and minimum S₁₁ when rotation angle θ_r swept from 0° to 360°) was ~2.5 and ~0 dB when L₂ was 2 and 15 mm, respectively. When ΔS₁₁ reduced to ~0 dB, this indicated that as the fan rotated, the S₁₁ was in fact kept almost constant. In other words, the rotation could not be detected in this case. These observations may be explained as follows: when L₂ increased, perturbations of the near-field of the microwave sensor caused by the fan blade decreased and, thus, S₁₁’s dependence on the position of the fan blades decreased.

Besides excitation frequency and distance L₂, the two alignment parameters Δ_x and Δ_y were also important for the sensor’s sensitivity. We ran a group of simulations by sweeping Δ_x and Δ_y to show their effect. The obtained simulation results are shown in Figure 7. We also used a three-blade aluminum fan here. Also, different from Table 1, in these simulations, L₂ was 5 mm. Other parameters were the same as those in Table 1. The simulated frequency band was also from 18 to 26 GHz, and, for simplicity, only the 22 GHz result is shown in Figure 7. It should be noted that, as shown in Figure 2d, the sensor was aligned with the fan when both Δ_x and Δ_y equaled zero. It can be seen from Figure 2 that when the sensor was misaligned with the fan to some degree, positioning the blade around the microwave spot, the perturbation variation of the sensor’s near-field by the rotating fan was maximized. This was verified by the simulation results shown in Figure 7. The width/height of a standard WR-42 waveguide is 10.7/4.3 mm, respectively. It can be seen that ΔS₁₁ reached its maximum when Δ_x and Δ_y were around 5 and 8 mm, respectively. From the sizes of the waveguide and fan, one can infer that these alignment parameters indicated an overlap between the fan blade and the microwave spot. Thus, the sensor’s sensitivity was maximized.

We ran a group of simulations using fans with different numbers of blades consisting of different materials. For demonstration, we present the results of fans with 3, 6 and 12 blades. Regarding the material, we used aluminum and alumina to represent metallic and non-metallic fans. The obtained simulation results are shown in Figure 8. Also, different from Table 1, in these simulations, L₂ was 5 mm, Δ_x was 10 mm and Δ_y was 5 mm. Other parameters were the same as those in Table 1. The simulated frequency band was also from 18 to 26 GHz, and, for simplicity, only the 26 GHz result is shown in Figure 8. It can be seen that when the rotation angle was swept from 0° to 360°, there were 3/6/12 periodicals for the fans with 3/6/12 blades. Namely, the sensing method was suitable for fans with various numbers of blades. However, as the number of blades increased, the amplitude variation of S₁₁ tended to decrease. For example, when the number of blades was 3, the amplitude variation of S₁₁ was ~3.5 dB, while it was ~0.5 dB for the fan with 12 blades. The dependence of the amplitude variation of S₁₁ on the number of blades may be attributed to the near-field distribution of the open-ended waveguide. Another observation shown in Figure 8 is that the proposed rotation sensor was also suitable for non-metallic fans such as the alumina fan. Thus, these results prove the feasibility of the sensor further in view of the number of blades and the fan’s material. Besides the open-ended waveguide, a waveguide horn is also a popular sensor and we are planning to try rotation sensing with a horn in the near future. Also, a dielectric-filled waveguide may also be used for this purpose.

3. Experiments

A schematic view of our first experimental setup is shown in Figure 9. An RF source (SynthHDPro V2, Windfreak Technologies, New Port Richey, FL, USA) was used for transmitting an 18 GHz continuous sinusoidal signal with 5 dBm of power. Here, the selection of working frequency was somehow arbitrary. In fact, we also tried with other frequencies and similar results were obtained. This signal then went through a circulator (UIYCC1318A16T18SF, UIY Inc., Shenzhen, China) and was fed into a waveguide-to-coaxial adapter (HD-220WCAK, HengDa Microwave, Xi’an, China). This adapter was used as a microwave sensor (the adapter could be seen as an open-ended waveguide) and it was able to emit part of the power fed into it into the space in front of it. However, the sensing method presented here mainly used its near-field energy around the waveguide’s mouth. Due to the rotating blade of the fan, the near-field microwaves were periodically perturbed, resulting in a time-dependent reflection coefficient of the sensor. Thus, the reflected microwave from the sensor was also time-dependent and the reflected microwave went into the RF detector (8472B, Keysight, Santa Rosa, CA, USA) through the circulator. The RF detector converted the received microwaves into DC (direct current) voltage, which was displayed on an oscilloscope (SDS 1202X-C, Siglent, Shenzhen, China). A photo of our measurement system as well as the fan and waveguide adapter are shown in Figure 10. We used a DC motor with a plastic three-blade fan as the rotating part. Its rotational speed could be adjusted by the excitation DC current supplied by the DC power source. It should be noted that to increase sensitivity, the blades of the plastic fan were metalized with adhesive copper tape (see Figure 10b). In fact, as shown later, the proposed near-field microwave method also works for non-metalized blades.

Figure 11 shows the typical voltage waveforms displayed on the oscilloscope for different motor currents. Since the number of blades was three (see Figure 10b), the periodicity of the voltage waveform was expected to be 1/3 of that of the rotating fan. It can be seen that the observed waveforms exhibited good periodicity. The periodicity of the waveform was measured manually using the measure line function of the oscilloscope, without any signal post-processing. However, to improve accuracy and realize automation, signal post-processing such as averaging, filtering and noise reduction could be used. In fact, the effect of signal post-processing using filtering will be shown later. One could measure a predefined number of periods and average the obtained rotational speeds corresponding to each single period. Another observed phenomenon was that the voltage amplitude was almost constant for different rotational speeds. This characteristic shall be helpful for signal post-processing.

In Figure 12, the measured rotational speeds of the fan when applying different motor currents are shown. The rotational speed was extracted according to the following procedure: first, the periodicity T in seconds (it needed to include three similar waveforms due to the number of blades being 3) of the waveform was read as shown in Figure 11; then, the rotational speed in RPM was calculated by dividing the periodicity T with 60 s, namely, 60/T, as shown by Equation (1). It can be seen that when the motor current increased from 0.25 A to 0.75 A, the measured rotational speed increased from ~1300 to 4800 RPM. For verification, we also used an optical transmission method, as schematically shown in the inset of Figure 12. A laser source emitted a focused optical beam (wavelength ~800 nm) that was shed on the blade from the metalized side of the fan. A photodiode was coaxially aligned with the laser but at the plastic side of the fan. The output of the photodiode was also monitored by an oscilloscope. As the fan rotated, it blocked the optical path periodically, which resulted in a similar waveform on the oscilloscope as the microwave method. It can be seen that the measured rotational speeds of the microwave method and the optical method agreed well with each other. The average difference of the measured rotational speeds was ~1.7%. We also used a commercially available digital tachometer (VC6234P, Victor Inc., China; it is based on the optical reflection method) for further verification, as shown in Figure 12. It can be seen that all of these three methods agreed well with each other.

One important aspect of the proposed measurement method was the amplitude of the voltage waveform. Besides the system design factors, such as the RF signal’s frequency/power and the waveguide/detector’s characteristics, the horizontal location (alignment between the fan and the microwave sensor) of the sensor relative to the fan was one of the most important factors influencing the observed voltage amplitude. As a preliminary evaluation, for various detection distances R (as shown in the inset of Figure 13, equivalent to L₂ shown in Figure 2), the x position (equivalent to Δ_y in the simulations) of the fan was adjusted and the corresponding voltage amplitudes were measured, as shown in Figure 13. The parameter R can be seen as the working distance of the microwave sensor. The maximum R used here was ~20 mm. It should be noted that when the x position was 0 mm, the center of the fan was aligned with that of the waveguide aperture. The following can be concluded: 1. The voltage amplitude increased with decreasing R. 2. It seems that when the center of the blade, instead of the fan, aligned with the waveguide aperture, a higher sensitivity could be achieved. Thus, there was an optimum value of x to achieve the highest sensitivity.

To further demonstrate the feasibility of the proposed rotational speed sensor, we used radiator fans for additional measurements. The measurement system used was similar to the one shown in Figure 10 and is shown in Figure 14. Here, we used another RF detector (QD-10-40000-N-K, Qualwave Inc., Chengdu, China). The rotation fan was fixed on an xyz positioning stage to control the fan’s position relative to the microwave sensor, which is helpful in studying the potential effect of alignment and distance. The radiator fan was non-metallic and adhesive copper tape was applied to some of the fan’s blades, as shown in Figure 14b.

A typical waveform obtained from the setup is shown in Figure 15a. In this measurement, the applied direct current voltage on the radiator fan was 13 V. The distance between the sensor and the fan was 2 mm and the location of the sensor was y = 25 mm and z = 25 mm. The definition of y and z is shown in Figure 14. The origin of the coordination was defined at the lower right corner of the fan. The excitation frequency was 18 GHz and the output power of the signal generator was 0 dBm. The number of metalized blades was 3 and the size of the fan was 120 mm (the edge length). It can be seen that the acquired data shown in Figure 15a were somehow noisy. So, we used a Gaussian-weighted moving average filter to smooth the data. After this post-process, we obtained the waveform shown in Figure 15b. From this post-processed waveform, one could easily extract its periodicity and thus finally obtain the fan’s rotation speed. As denoted in Figure 15b, each blade of the fan contributed to the fluctuation of the voltage waveform, and the metalized blades generated a more obvious variation of the output DC voltage. In other words, both the metalized and non-metalized blades of the radiator fan could be detected by the microwave sensor. In Figure 16, the sensor’s response under various distances between the sensor and the fan is shown. When the distance ranged from 2 to 17 mm, periodical waveforms could be obtained, indicating their ability to sense rotation.

The simulation results shown in Figure 4 indicate that the excitation frequency and the distance between the sensor and the fan should be carefully considered to achieve reasonable sensitivity. To demonstrate this experimentally, a group of measurements were conducted. In these measurements, experimental parameters were the same as those in Figure 15. Three excitation frequencies were used, namely, 17.6, 17.8 and 18 GHz. The distance between the sensor and the fan was swept from 2 to 20 mm. Similar to the simulation results mentioned above, the acquired data shown in Figure 17 demonstrate that both the excitation frequency and distance influenced the observed amplitude of the voltage waveform. Here, ΔV was defined as the maximum variation of a voltage waveform, as shown in Figure 16. Regarding the effect of excitation frequency on sensitivity, this could be attributed to the near-field distribution as well as the frequency response of system components like the RF detector and circulator. Regarding the distance, generally speaking, the sensitivity reduced as the distance increased. This is easy to understand since the field intensity decreased with increasing distance.

Figure 18 shows the observed effect of excitation power on the sensor’s performance. In these measurements, experiment parameters were the same as those in Figure 15. The excitation frequency was 18 GHz and the distance between the sensor and fan was 5 mm. It can be seen that the observed amplitude of the voltage waveform increased with excitation power. This could be attributed to the characteristics of the RF detector used. A higher output power of the signal generator induced a higher input power of the RF detector. Due to its logarithmic transfer characteristic, the output voltage of the RF detector increased with input power. So, increasing the excitation power may be a simple way to increase the sensitivity of the rotation system.

The simulation results shown in Figure 7 indicate that the location of the sensor relative to the rotating fan influenced the sensitivity and it seems that, for a given fan size, there exists an optimum relative location to achieve the highest sensitivity. As shown in Figure 19, to demonstrate this experimentally, a group of measurements were conducted. In these measurements, experimental parameters were the same as those in Figure 15. The excitation frequency was 18 GHz. The distance between the sensor and the fan was 5 mm. The locations of the sensor, namely, y and z, were swept from −125 to 20 mm. It can be seen that, similar to the simulation results mentioned above, the observed amplitude of the voltage waveform changed with y and z. Furthermore, the experimental results also show that the sensitivity reached its highest value for some intermediate locations. If the fan was located out of these intermediate locations, the sensitivity reduced sharply or even failed to work (the output voltage amplitude became lower than the system’s noise floor). From a practice point of view, this observation indicates that a sensor’s installation location should be considered to achieve reasonable performance.

Finally, a group of measurements were conducted to preliminarily show the effect of the number and material of blades as well as the size of the fan. In these measurements, experimental parameters were the same as those in Figure 15. The excitation frequency was 18 GHz. The distance between the sensor and the fan was 2 mm. The location of the sensor was y = −25 mm and z = −25 mm. Three fans were used: the first one was the same as that shown in Figure 14; the second one was a fan with non-metalized blades and its size was 90 mm; the third one was a fan with one metalized blade (other blades were non-metalized) and its size was 60 mm. The obtained results are shown in Figure 20. It can be seen that all of the three cases presented periodical waveforms, indicating the possibility of rotation sensing. Thus, these results show the potential for the wide application of the proposed rotational sensor.

4. Conclusions

A rotational speed measurement method based on a near-field microwave using an open-ended waveguide was presented in this work. Simulations and experiments were presented for verification. A commercially available waveguide-to-coaxial adapter working at 18 GHz was used as a sensor. Measurement results of a plastic fan with both metalized and non-metalized blades prove that the proposed method operates well for rational speeds of up to ~5000 RPM. The proposed method was also compared with the optical transmission/reflection method, which shows 1.7% accuracy. Future work may include the use of lower frequency band waveguides or planar transmission lines, such as substrate-integrated waveguides, and system integration for practical usage.