This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

Optimized Differential Self-Inductance

Displacement Sensor for Magnetic Bearings:
Design, Analysis and Experiment
Kun Wang, Lisheng Zhang, Yun Le, Shiqiang Zheng, Bangcheng Han and Yinxiao Jiang

suspended and rotated at a high speed; (2) The detection range

Abstract—The miniaturization requirements have been put of the displacement sensor can be adjusted according to the
forward to the sensors for the development of the modern different operation condition, and the desired detection range
industry. However, the traditional inductive displacement sensor for AMBs system is 0.5 ~ 1.0 mm; (3) The satisfactory static
applied to active magnetic bearings (AMBs) has complicated
detection device and measuring circuit with excessive discrete and dynamic characteristics, which means a sensitivity slightly
components which cannot meet the requirements of the larger than 20 mV/um, a linearity as high as possible and a
miniaturization. To remedy the infliction, an optimized resolution of 0.1 um or even smaller, should be realized. For a
differential self-inductance displacement sensor was designed for AMBs system with a 60000 rpm rated speed, the bandwidth of
AMBs. Concretely, the operating principle of the self-inductance the sensor should be no less than 1 kHz. And the wider the
displacement sensor was analyzed in detail. Then the optimized bandwidth of the sensor is, the better the dynamic performance
structure was proposed and the measuring circuit was designed.
After the designing, the performance of the sensor was analyzed. of the AMBs control system can be achieved; (4) High capacity
Based on the analysis and designing, a series of experiments were of resisting disturbance [10]. At present, the eddy current
carried out to test the performance. The results demonstrated that displacement sensor can meet the AMBs requirements in most
the designed sensor can realize higher linearity, higher sensitivity situations and is widely used by majority prototypes for its easy
and wider dynamic response range than the traditional inductive implementation [11-12]. However, the eddy current
displacement sensor. And it is suitable to the AMBs system, with displacement sensor can be easily affected by the external
higher rotation speed and small volume.
magnetic field when the maglev rotor rotated at an ultra-high
Index Terms—magnetic bearings, self-inductance displacement speed. And the signal-to-noise ratio (SNR) of the sensor will be
sensor, performance, miniaturization reduced drastically by the inherent noises caused by the
converting circuit [13].
The inductive displacement sensor possesses all the
I. INTRODUCTION advantages of the eddy current displacement sensor except for
the lower frequency response. Besides, it has the advantages of
A S a new kind of supporting mechanism, the AMBs
supports rotating rotor without any mechanical contact by
electromagnetic forces, and basically eliminates the friction
high SNR and high capacity of resisting disturbance. If the
frequency response of the inductive displacement sensor is
caused by the mechanical ball bearings [1-4]. With the properly compensated, it can reach to a wider application
outstanding merits of no lubrication, low power consumption, prospect than the eddy current displacement sensor in the
long life, high precision and no pollution to the environment, high-precision AMBs system [14-19]. Therefore many
the AMBs was widely used in the field of space technology and researchers had deeply studied about the inductive
other high-tech fields, such as air blower, gas compressor, displacement sensor.
molecular pump, flywheel in the satellite, gyroscope, machine In the literatures [20], the radial displacement of the maglev
tool spindles, and so on [5-7]. It is extensively applied in the rotor was measured by the differential combination of the
industrial field of manufacturing, energy, military and inductive displacement sensors. Unfortunately, this gave rise to
aerospace [8]. The AMBs can keep the rotor position precisely a remarkable systematic error because the differential
according to the different operation state by adjusting the procedure was done in the subsequent signal processing. To
electromagnetic force. To realize the high precision suspended solve this problem, the differential transformer type
of the rotor, a displacement sensor with high sensitivity and displacement sensor and the dedicated chip AD698 produced
resolution is required to detect the rotor position in real time. by the American company AD for signal conditioning were
Therefore, the high-precision displacement sensor becomes a adopted in paper [21]. But the output excitation frequency of
crucial component for AMBs to complete the robust distance the AD698 can only vary from 20 Hz to 20 kHz. And the cut-off
detection of the high-speed maglev rotor [9]. frequency of the processing circuit was only 800 Hz which
There are 4 requirements of the AMBs for the displacement cannot meet the requirements of the high-speed AMBs. In [22],
sensor: (1) The displacement of the rotor position should be the differential rectifier circuit and the phase-sensitive
detected in real time because the rotor needs to be stably detection circuit were proposed to overcome the disadvantages

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

of the AD698. However, the rectifier characteristics were not magnetic field in the air gap is uniform when the air gap is
ideal due to the threshold voltage of the diodes in the extremely small. Ignoring the hysteresis losses, the total
differential rectifier circuit which employed a host of diodes. magnetic reluctance can be described as follows:
And the determination of the polarity of the output voltage  l  2δ 1 l l 2δ 1
required the use of the primary excitation voltage as a phase Rµ = ∑  i  + = 1 + 2 + (2)
µ A µ µ µ µ
reference when using the phase-sensitive detection circuit. In  i i A
0 0 1 1A A
2 2 0 A0

[23], the inductive displacement sensor using a differential where li is the length of each section of the magnet. μi is the
inductive sensor probe, lock-in amplifier, low-pass filter and permeability of each section of the magnet. Ai is the sectional
notch filter was implemented. As a consequence, the area of each section of the magnet. δ1 is the thickness of the air
satisfactory linearity, the high sensitivity and the eliminated gap. μ0 is the vacuum permeability. A0 is the sectional area of
disturbances which caused by the natural frequencies of the the air gap. Generally, the magnetic reluctance of the air gap is
rotor were achieved by the sensor. Whereas that the much larger than the magnetic reluctance of the iron core and
requirements of the industrial miniaturization were not satisfied the armature.
because of the use of a multitude of chips. Therefore, the design  2δ 1 l1
of the inductive displacement sensor with the superior µ A 2 µ A
 0 0 1 1
performance and the simplified measuring circuit can make  (3)
 2δ 1 l2
great significance to the AMBs system [24]. To this end, an 2
optimized differential self-inductance displacement sensor with  µ0 A0 µ2 A2
simplified structure and measuring circuit was designed in this Then the Rm can be calculated as:
paper. 2δ 1
The paper is organized as follows. Section II introduced the Rµ = (4)
µ0 A0
operating principle of the self-inductance displacement sensor
The inductance of L1 can be described as follows by the
in detail and analyzed the performance of the sensor. To further
formulas (1) and (4):
improve the performance, an optimized structure was proposed
and analyzed in Section III. Section IV presented the designing µ0 A0W 2
L1 = (5)
of the measuring circuit. And a series of experiments were 2δ 1
carried out to test the sensor performance in Section V. The Similarly, the inductance of L2 can be obtained.
conclusions can be found in Section VI. µ0 A0W 2
L2 = (6)
2δ 2
II. PRINCIPLE ANALYSIS
As shown in the formulas (5) and (6), there is a nonlinear
The rotor displacement can be converted to the variation of relationship between the inductance L and the air gap δ. The
the self-inductance or the mutual inductance in the coils based characteristic is shown in Fig.2.
on the principle of electromagnetic [25]. Then the variation of ∆L
the self-inductance or the mutual inductance is converted to the L1 C1
B
available voltage or current for AMBs by the measuring circuit. C

In this paper, the differential inductive sensor was designed

using the self-inductance principle. The differential structure of
the sensor mainly consists of coil, iron core and armature as 0 ∆δ
Fig.1 shows.
l1 Coil
L1

A1 Iron core
L2
W
Linear range
l2 Fig.2. L-δ characteristic of the differential inductive sensor.
δ1
The equivalent circuit of the differential inductive sensor is
δ2 shown in Fig.3.
Armature

A2 Us L1

L2
Fig.1. Differential structure of the sensor. 2Us Uo

The inductance of L1 can be obtained by follow equation:

W2 Us L2
L1 = (1)
Rm
where W is the number of turns of the coil. Rm is the total Fig.3. Equivalent circuit of the differential inductive sensor.
magnetic reluctance. It is universally acknowledged that the Then the output voltage can be calculated as follows:

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

L2 − L1 The finite element model of the sensor was established as

Uo =
Us (7) Fig.6 shows. Where Fig.6 (a), Fig.6 (b), Fig.6 (c) and Fig.6 (d)
L1 + L2
are the distribution maps of the magnetic lines within the
According to the formulas (5) and (6), and the equations δ1=
probes with the moving distances of the rotor are 0 mm, 0.1 mm,
δ0 + ∆δ and δ2 = δ0 - ∆δ, the output voltage can be gained as
0.2 mm and 0.3 mm respectively. It can be concluded from
follows:
Fig.6 that the magnetic field in the sensor probes is symmetrical
∆δ
Uo = Us (8) when the rotor is suspended in the center and there is almost no
δ0 magnetic flux leakage and mutual interference in the probes.
A[Wb/m] A[Wb/m]
The Eq. (8) demonstrates that the output voltage is linear
1.6734e-005 2.2431e-005
with the rotor displacement no matter the rotor moves up or
2.0069e-005
down. So the displacement of the rotor can be acquired in the 1.4500e-005
1.7707e-005
condition of that the change of the output voltage can be 1.2265e-005 1.5344e-005
detected. It is inevitable to improve the inductive displacement 1.0030e-005 1.2982e-005
sensor to complete the more harsh distance detection for 7.7955e-006 1.0619e-005
AMBs. 5.5608e-006
8.2570e-006
5.8946e-006
3.3261e-006
III. OPTIMIZED DESIGN 3.5322e-006
1.0914e-006 1.1698e-006
A. Improved Structure Design (a) (b)
A[Wb/m]
In order to detect the displacement of the rotor more 3.5140e-005
A[Wb/m]
8.9096e-005
sensitively, the structure was improved as Fig.4 shows. The coil 3.2704e-005
8.1769e-005
3.0268e-005
P1 is connected in series with the coil P2. The coil N1 is 2.7831e-005 7.4443e-005
2.5395e-005 6.7117e-005
connected in series with the coil N2. And the numbers of turns 2.2959e-005 5.9790e-005
of the four coils are equal. 2.0523e-005
1.8087e-005
5.2464e-005
4.5137e-005
1.5651e-005 3.7811e-005
1.3215e-005 3.0484e-005
1.0778e-005
P2 8.3422e-006 2.3158e-005
P1
5.9061e-006 1.5831e-005
3.4700e-006 8.5051e-006
δ 1.0338e-006 1.1786e-006
��1
δ2 (c) (d)
Fig.6. Distribution maps of the magnetic lines within the probes. (a) Rotor
N2 N1
displacement of 0 mm. (b) Rotor displacement of 0.1 mm. (c) Rotor
displacement of 0.2 mm. (d) Rotor displacement of 0.3 mm.
The inductance variations were calculated when the rotor
Fig.4. Improved structure of the differential self-inductance displacement moved towards XP. The result is shown in Fig.7.
sensor. -0.2 0.0 0.2 0.4
100
The requirements of the sensor cores on the magnetic circuit,
80
the travel stroke and the non-linearity of the sensor should all be
60
taken into account in the magnetic structure design. The 40 Single
G (XP1)
magnetic structure was designed not only to make the sensor 100

have enough core window area and heat dissipation area, but 80
60
also considering the characteristics of the AMBs structure and Single (XN1)
40 F
the space constraints as shown in Fig.5. 250
200
Inductance (mH)

150
YP2- 100 Pair
E (XP1, XP2)
XP1+
250
200
YP2+
XP1- 150
YP1+ S S XP2-
YP1- XP2+ 100 Pair
D(XN1, XN2)
N N 200
150
100 Unimproved
XN2+
N N YN1- C (L1)
S S 200
XN2- YN1+
YN2+ 150
XN1-

100 Unimproved
XN1+ YN2-
B (L2)
50
-0.2 0.0 0.2 0.4
Displacement (mm)
Fig.5. Magnetic structure of the differential self-inductance displacement Fig.7. Inductance varies with the rotor displacement.
sensor with coils. As shown in Fig.7, the inductances of single coil XP1, pair

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

coils XP1 and XP2 and unimproved coil L1 increase when the 48 finally.
rotor moves towards XP. The inductances of single coil XN1, pair
C. Excitation Frequency Analysis
coils XN1 and XN2 and unimproved coil L2 decrease when the
rotor moves away XN. L1 and L2 are calculated according to the The inductance and the internal resistance of the sensor vary
previous structure. And the number of turns of the coils L1 and with the excitation frequency. And the linearity of the sensor
L2 is twice that of the single coil XP1. Coil L1 is in the XP can be affected by the ratio of the internal resistance to the
direction. Coil L2 is in the XN direction. The difference in the inductive reactance. Fig.9 shows the frequency characteristics
inductance variation ∆L of the symmetrical coils is shown in of the sensor.
Fig.8.
200
200
∆L
100 190
190

180
180
80

Inductance (mH)
Single (XP1, XN1) 170
170
Pair (XP, XN)
60 Unimproved (L1, L2) 160
160

150
150
40
140
140

130
130
-0.2 -0.1 0 0.1 0.2 ∆δ
-40 120
120

110
110
-60 00
0.1
0.1 00
Disp 0.2
0.2
20
10
1
lacem 0.30.3
30
3
2
-80 ent (m 0.4
0.4
m) 40
4 cy (kHz) 4
0.5
0.5 50
5 Frequen x 10

-100 (a)
Fig.8. ∆L-∆δ characteristic.
3500
3500
From Fig.8, it can be seen that the improved structure
enhances the linearity and increases the linear range of the 3000
3000

sensor. 2500
Internal resistance (Ω)

2500

B. The Number of Turns of the Coils 2000

2000

The iron core lamination is consisted of silicon steel sheet 1500

1500
with thickness 0.2 mm. The diameter of the enameled wire is
1000
1000
0.25 mm. The length of the skeleton is 3 mm. The following
formulas hold according to the current density and the window 500
500

area for the coils: 00

0.5
0.5
0.4
I 0.4 5
jg =
0.3
0.3 4 50
(9) Displa 0.2
ceme 0.2 0.1 2
20
3
30
40
q nt (m 0.1
m) 00 00
1
10
Frequency (k
Hz)
x 10
4

Qkc (b)
q≥ (10)
N
where jg is the current density. I is the maximum carrying 1.4
1.4
Internal resistance / inductive reactance

capacity of the enameled wire. q is the sectional area of the wire. 1.2
1.2
Q is the core window area of the skeleton. kc is the filling factor 1.01
of the window winding and kc = 0.7. N is the number of turns of 0.8
0.8
the coils. 0.6
0.6
In order to ensure that the coil is not overheated and burned:
0.4
0.4
Am ≥ I 2 Rc An (11) 0.2
0.2

where Am is the surface area of the coils. Rc is the internal 00

resistance of the enameled wire. An is the area of heat 00
0.1
0.1
0.2 1
00
dissipation per Watt. 0.2 2
10
Disp 0.3
0.3 20
lacem 3
30
Considering the nonlinear distortion of the small signal, the ent ( 0.4
0.4 4
40 (kHz) 4
mm) 0.5
0.5 5
50 Fre quency x 10

magnetic induction B in the air gap is generally lower than the (c)
saturation point to ensure that the B is in the linear segment, Fig.9. Frequency characteristics of the sensor. (a) Inductance varies with the
frequency. (b) Internal resistance varies with the frequency. (c) The ratio of the
where Bc = 0.75 T, then internal resistance to the inductive reactance varies with the frequency.
µ0 Hδ ≤ Bc (12) As shown in Fig.9 (a), the inductance of the sensor decreases
where Hδ is the maximum magnetic field strength in the air gap. as the frequency increases. As shown in Fig.9 (b), the internal
Bc is the magnetic induction of saturation point on the basic resistance of the sensor increases as the frequency increases. As
magnetization curve. shown in Fig.9 (c), the ratio of the internal resistance to the
We considered the constraints of the parameters and took N = inductive reactance decreases at first, then increases and finally

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

decreases again as the frequency increases. The ratio of the The chip ICL8038 is a precision voltage control function
internal resistance to the inductive reactance was too large and generator developed by the INTERSIL with the characteristics
did not meet the actual situation obviously when the excitation of high stability, high accuracy and low output waveform
frequency was less than 1 kHz. So ignoring the values that did distortion. The frequency of the excitation signal is determined
not meet the actual working situation, the ratio of the internal by Ra1, Ra2, Ca1 and Ca2. RV2 is used to adjust the distortion of
resistance to the inductive reactance can be obtained as shown the excitation signal.
in Fig.10.
B. Envelope Detection Circuit Design
The high-frequency carrier signal in the amplitude
Internal resistance / inductive reactance

0.095
0.095
modulation wave can be filtered out by the demodulator circuit,
0.09
0.09
which can detect the high-frequency amplitude modulation
0.085
0.085
signal to obtain the actual measured signal. The envelope
0.08
0.08
detection circuit was shown in Fig.13.
0.075
0.075 Db1 Rb1
0.07
0.07

0.065
0.065

0.06
0.06
0 Cb1 Rb2
0.1
Dis
pla 0.2
0.2
cem 0
ent 0.3
0.3 10
1
(m 20
2
m) 0.4
0.4 30
3
40 (kHz) 4
Frequency
4 x 10
0.5
0.5 50
5
Fig.13. The envelope detection circuit.
Fig.10. The ratio of the internal resistance to the inductive reactance varies with
the frequency more than 1 kHz. The detection circuit consists of a charging loop and a
From Fig.10, it can be seen that the minimum ratio of the discharging loop. The charging loop is composed of the
internal resistance to the inductive reactance happens at the detection diode Db1, the positive on-resistance Rdb1 and the
excitation frequency of 25 kHz. So the excitation frequency of detection capacitor Cb1. The discharging loop is composed of
25 kHz was adopted. the detection load resistances Rb1, Rb2 and the detection
capacitor Cb1. Obviously, the time constant τ2 of the discharge
IV. DESIGN OF THE MEASURING CIRCUIT circuit is much larger than the time constant τ1 of the charging
circuit. The model of the detection circuit can be equivalent to
In order to obtain the information of the rotor displacement,
two first-order inertia links in series. The two links represent
the output amplitude modulation wave must be subjected to
the charging circuit and the discharge circuit, respectively. The
demodulating, differential amplifying, and filtering, etc. Fig.11
transfer function of the detection circuit is:
shows the detecting principle of the sensor.
1 1 1 1
= G1 ( s ) = (13)
τ1s + 1 τ 2 s + 1 Rdb1Cb1s + 1 ( Rb1 + Rb 2 ) Cb1s + 1
In the actual circuit, the resistance of Rdb1 is 100 Ω, Rb1 and
Rb2 are 200 kΩ. The capacitance of Cb1 is 150 pF. These values
are substituted into the formula (13).
1 1
G1 ( s ) = (14)
1.5 × 10 s + 1 6 × 10−5 s + 1
−8
Fig.11. The detecting principle of the sensor.

A. Excitation Circuit Design C. Differential Amplifier Circuit Design

Fig.12 shows the circuit of sinusoidal excitation signal Fig.14 shows the adopted differential amplifier circuit
generated by the ICL8038. composed of 3 operational amplifiers.
+12V
+12V
Vin1 50R
%c7
Rc3 Rc5

Ra4 Ra5 66 Rc1

-12V -12V
88 44 Ra1
77 55 Ra2
50 %
RV1 Vout
11
Rg
50RV
%2
12
12 99 Ra3
10
10

33 22 Excitation
-12V Rc9
Rc2 +12V -12V
11
11
50R%
C a1 Ca2 ICL8038 Rc4 Rc6 c8

Vin2
Rc10
+12V

+12V
-12V
Fig.12. Generation circuit of the sinusoidal excitation signal. Fig.14. The differential amplifier circuit.

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

In order to keep the current loop of the sensor safety, the components are taken as:
output impedance of the operational amplifier has to be Rd1 = 1.2 kΩ; Rd2 = 2.4 kΩ; Rd3 = Rd4 = 8.2 kΩ; Cd1 = Cd2
increased. The differential amplifier circuit includes the =0.02 μF.
amplification function of differential input and single-ended The transfer function of the filter circuit is:
output. And Rc1 = Rc2, Rc3 = Rc4, Rc5 = Rc6, Rc9 = Rc10. Rc7 was 1.74 × 109
used to adjust the output error. Rc8 was used to adjust the G2 ( s ) = (21)
s 2 + 4.17 × 104 s + 8.68 × 108
reference voltage.
The output voltage of the differential amplifier circuit is: The responding amplitude - frequency characteristic and the
phase-frequency characteristic can be obtained and shown in
 2 Rc1  Rc5 
Vout = (Vin 2 − Vin1) 1 +   Fig.16.
Rg   Rc3 
(15)
 2.5
2.5

System: sys
Using the differential amplifier circuit, the required 22 Frequency (Hz): 4.71k
Frequency (Hz): 4.75e+003

(abs)
Magnitude(abs)
Magnitude (abs): 1.414
Magnitude (abs): 1.4
front-end gain (determined by the Rg) can be obtained without 1.5

Magnitude
increasing the common-mode gain and error. 11

D. Low Pass Filter Design 0.5

The filtering circuit was used to remove the high-frequency 00

00
wavebands of the output from the detection circuit, and to
obtain a consistent rotor displacement signal. So the -45
-45

(deg)
Phase(deg)
Butterworth filter was adopted because its amplitude-frequency -90
-90

Phase
characteristic in the pass band is the closest to the ideal situation. -135
-135
Fig.15 shows the Voltage-Controlled Voltage Source (VCVS)
-180
-180
circuit of the Butterworth filter. 10
3 4
104 10
10
5
5

Cd2 Frequency (Hz)

F (H )
Fig.16. The responding amplitude - frequency characteristic and the
-12V phase-frequency characteristic of the filter.
Rd1 Rd2
E. Compensation Circuit Design
The dynamic response characteristic of the inductive sensor
Cd1 +12V designed in this paper is mainly determined by the discharge
Rd4 loop of the detection circuit and the frequency characteristic of
the filter circuit. The series of the above steps can be used to
Rd3
represent the equivalent transfer function of the sensor:
1 1 1.74 × 109 (22)
=G ( s ) G=
(s)G (s)
1.5 × 10−8 s + 1 6 × 10−5 s + 1 s 2 + 4.17 × 104 s + 8.68 × 108
1 2

Fig.15. The VCVS circuit of the Butterworth filter. To improve the response bandwidth of the sensor, it is
The transfer function of the VCVS circuit is: indispensable to reduce the time constant of the discharge loop
1 and increase the cut-off frequency of the filter circuit. But to
Av
Rd 1 Rd 2Cd 1Cd 2
G2 ( s ) = reduce the time constant of the discharge loop, the detection
 1 1 1  1 capacitor and the detection resistance should be reduced. The
s2 +  + + (1 − Av ) s+
 Rd 1Cd 1 Rd 2 Cd 2 Rd 2 Cd 1  Rd 1 Rd 2 Cd 1Cd 2 reduced load will cause output distortion, increase the ripple
(16) noise and reduce the SNR of signal. Increasing the cut-off
The G2(s) can be rewritten as: frequency of the filtering circuit will reduce the low-pass
performance and increase the high-frequency noises of the
Avωc2
G2 ( s ) = (17) output. Therefore, a cascade leading compensator to correct the
ωc
s2 + s + ωc2 frequency characteristic and improve the dynamic response
Q characteristic of the sensor was adopted as shown in Fig.17.
where Av is the pass-band gain. ωc is the cut-off frequency of the Ce1
filter. Q is the equivalent quality factor.
Re1
R
Av = 1 + d 4 (18)
Rd 3 Re2

1
ωc = (19) Fig.17. The passive advanced network.
Rd 1 Rd 2 Cd 1Cd 2
The transfer function of the passive advanced network is:
Rd 1 Rd 2 Rd 3Cd 1Cd 2 Ts + 1
Q = ωc (20) Gc ( s ) =
(23)
Rd 1 Rd 3Cd 1 + Rd 1 Rd 3Cd 2 − Rd 1 Rd 4 Cd 2 α Ts + 1
The parameters of the components are designed according to where α < 1. ω1 = 1 / T and ω2 = 1 / (αT) are turning frequencies.
the cut-off frequency of 4 kHz. The standard values of the Based on the equation (23), it can be obtained:

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

∠Gc ( jω )= arctan ωT − arctan aωT (24) The output voltage is detected based on the variation of rotor
displacement from -0.5 mm to 0.5 mm by step length of 0.05
There is a maximum lead angle φm between the turning
mm. And the arithmetic mean of the output voltage is obtained
frequencies ω1 and ω2, and the extremum comes from the
by 6 repeating and independent experiments. The linear fitting
equation (24) when ω = ωm = 1 / (α1/2T)
curve of the experimental data is shown in Fig.20. Fig.21 shows
1 − sin ϕm
α= (25) the residual curve. Then the performance indicators of the
1 + sin ϕm sensor can be obtained: linearity, sensitivity, hysteresis,
where the maximum lead angle φm occurs just at the geometric repeatability, as shown in Table I.
center of the two turning frequencies ω1 and ω2.
15
According to the requirement of the dynamic response
bandwidth of the sensor, α = 0.1585 and T = 0.0000646 are 10

Output (V)
obtained. The transfer function of the designed passive 5

advanced network can be gained by the equation (23): 0

0.0000646 s + 1 -5
Gc ( s ) = (26) 6 Experiment
0.00001024 s + 1 -10
Linear fit
Based on the basic formulas α = Re2 / (Re1+Re2) and T = Re1Ce1. -15

Taking Ce1 = 2.2nF, we can get Re1 = 29.4 kΩ, Re2 = 5.54 kΩ. -0.5 -0.25 0.0 0.25 0.5
The standard values are taken that Re1 = 30 kΩ, Re2 = 5.6 kΩ. Displacement (mm)
Fig.18 shows the comparison of the frequency characteristics Fig.20. Volt-Displacement characteristic.
of the sensor before and after compensation. The cut-off Maximum residual
0.3
frequency increases from 2.44 kHz to 4.67 kHz. The using of point (0.5, 0.33188)

Residual of output (V)

the compensation circuit greatly improved the dynamic 0.15
response characteristics of the sensor.
g
20
20 0.0
00
(dB)

System: 补偿前 System: 补偿后 -0.15

Magnitude(dB)

-20
-20 Frequency (Hz): 2.44k Frequency (Hz): 4.67k
Frequency (Hz): 2.45e+003 Frequency (Hz): 4.65e+003
Magnitude (dB): -3 Magnitude (dB): -3
Magnitude

-40 Magnitude (dB): -3.04 Magnitude (dB): -2.96

-40
-0.3
-60
-60

-80
-80 -0.5 -0.25 0.0 0.25 0.5
Displacement (mm)
-100
00 Fig.21. The residual curve.
补偿前
Before compensation TABLE I
补偿后
After compensation PERFORMANCES OF THE SENSOR
(deg)
Phase (deg)

-90
-90
Measuring Linearity Sensitivity Hysteresis Repeatability
Phase

range(mm) (%) (mV/um) (%) (%)

-180
-0.50~0.50 ±1.25 25.94 ±0.28 ±0.22
-270 Compared with the performance of the traditional sensors
2 33 44 55
10 10 10 10
10 10
Frequency (Hz)
(Hz)
10 10 [22][26][27], the linearity and the sensitivity of the sensor we
Frequency
Fig.18. The comparison of the frequency characteristics before and after proposed have been greatly improved in the same detection
compensation. range of -0.5 mm~0.50 mm. At the same time, the hysteresis
and the repeatability reduced to a certain extent. With the better
V. EXPERIMENT EVALUATION performance than the conventional inductive sensor, it is more
A. Static Characteristics Test appropriate for AMBs control system.
The static characteristics test of the designed sensor was B. Dynamic Characteristics Test
carried out on the static calibrator through the measuring circuit. The dynamic characteristics of the sensor are tested by a
Fig.19 shows the experimental setup. special setup as shown in Fig.22. The setup mainly includes the
±15V DC
oscilloscope, ±15 V DC power, lathe, dynamic calibration disc,
Oscilloscope
power supply and fixed probe bracket. The lathe was fixed with a dynamic
Sensor
calibration disc. The two symmetrical sensor probes were fixed
at equal distances from the calibration disc by the two fixed
probe brackets. There evenly distributed 63 grooves on the
calibration disc which rotated with the lathe at different speeds.
Measuring Then the displacement signal that the sensor probes detected
circuit board
can be generated by the groove of the calibration disc. The
Sensor maximum frequency of the lathe is 150 Hz. So the dynamic
Micrometer
characteristics test can be conducted in the frequency range of
Base
0-9 kHz. The dynamic characteristics curve of the sensor before
Fig.19. The experimental setup for the test on static characteristics. and after compensation is shown in Fig. 23.

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

±15V DC paper [26], which can greatly improve the controllability for
power supply
AMBs system.
1 500mV/ 2 500mV/ 3 500mV/ 4 500mV/ 0.0s 100.0ms
Oscilloscope Calibration
Pk-Pk(1): 30 mV
disc
Pk-Pk(2): 30 mV

Pk-Pk(3): 90 mV
Fixed

Voltage
M
bracket 1 Pk-Pk(4): 90 mV
Lathe

2
Fig.22. The experimental setup for the test on dynamic characteristics.
B 3
66 Before compensation
After compensation 4
55
time
44
Fig.25. The experimental result of the suspension test.
C1 (V)
Output

33
VI. CONCLUSION
22
In this paper, an optimized self-inductance displacement
11
sensor for AMBs with simplified structure and measuring
00 circuit was proposed and designed. The main conclusions are as
00 1
1000 2
2000 3
3000 4
4000 5
5000
Frequency
A (kHz) follows:
Fig.23. The dynamic characteristics curve of the sensor before and after (1) The new requirements of sensor for AMBs were
compensation. discussed. The sensor requires a more pleasurable performance
It can be concluded from Fig. 23 that the amplitude of the to detect the rotor displacement more precisely, and
dynamic characteristics of the sensor in the high-frequency streamlined measuring circuit to satisfy the industrial
band is reduced. But the amplitude of the dynamic miniaturization. To meet the requirement, an optimized
characteristics before compensation is not stable. It has been self-inductance displacement sensor was recommended for its
reduced to 4 V below the input frequency of 2 kHz. However, advantage of high performance and simplified measuring
the amplitude after compensation is attenuated by 3 dB when circuit.
the frequency is above 4.7 kHz and is basically no attenuation (2) In the self-inductance displacement sensor, the variation
in the frequency range of 2 kHz. Compared with the sensor of the inductance has a paramount influence on the linearity and
developed by the previous research [23][26], the optimized sensitivity of the sensor. To further improve the performance,
sensor we proposed can provide better dynamic performance the probe structure composed of two coils in series was
for AMBs system. considered.
C. Suspension Test (3) The excitation frequency can affect the linearity and
Fig.24 shows the experimental setup for the suspension test. sensitivity of the sensor. Consequently, the effect of a variety of
The molecular pump has the following characteristics. The different excitation frequencies were compared and imposed in
caliber is 160mm. The rated speed is 36000rpm. The pumping the optimal design.
rate is 700 L/s. The ultimate vacuum can reach to 10-8Pa. (4) In the new requirements, the components of the
measuring circuit should be simplified and the bandwidth
Oscilloscope should be much wider. Accordingly, the circuit module consists
of envelope detection circuit, differential amplifier circuit and
48V DC power
compensation circuit was discussed and applied in the
supply
±15V DC power
optimized design.
Emulator
supply (5) According to the analysis results, an optimized
self-inductance displacement sensor was designed, fabricated
and tested. The experimental results demonstrate that the sensor
Switch can realize higher linearity, higher sensitivity and wider
Controller Magnetic levitation
molecular pump dynamic response range, and meet the requirements of the
industrial miniaturization.
Fig.24. The experimental setup for the suspension test. The sensor design method presented in this paper is useful
Fig.25 shows the experimental result. The four-way and convenient for the application of the high-speed maglev
full-range output of the sensor is 12.5V / mm. The peak-to-peak rotating machinery in industry.
value of the four-way radial beating is 0.03 V, 0.03 V, 0.09 V,
and 0.09 V, respectively. So the suspension accuracy of the REFERENCES
magnetic suspension molecular pump reached 2.4 um when [1] L. Zhou and L. Li, “Modeling and identification of a solid-core active
applying the designed sensor. The suspension accuracy of the magnetic bearing including eddy currents,” IEEE/ASME Trans.
Mechatronics, vol. 21, no. 6, pp. 2784-2792, Dec. 2016.
sensor increased by 60% compared with the sensor proposed in

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2017.2710135, IEEE Sensors
Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

[2] Z. Huang, J. Fang, X. Liu and B. Han, “Loss calculation and thermal bearing test rig,” Proc. of the 34th Int. Spr. Sem. on Electronics
analysis of rotors supported by active magnetic bearings for high-speed Technology, 2011, pp. 626-631.
permanent-magnet electrical machines,” IEEE Trans. Ind. Electron., vol. [25] Y. Tao, W. Yin, W. Zhang, C. Ktistis and A. J. Peyton, “A very-low-
63, no. 4, pp. 2027-2035, Apr. 2016. frequency electromagnetic inductive sensor system for work-piece
[3] G. Liu and K. Mao, “A novel power failure compensation control method recognition using the magnetic polarizability tensor,” IEEE Sensor J.,
for active magnetic bearings used in high-speed permanent magnet motor,” DOI 10.1109/JSEN. 2017. 2676460.
IEEE Trans. Power Electron., vol.31, no. 6, pp. 4565-4575, Jun. 2016. [26] C. Jin, L. Xu and J. Zhou, “Application of differential transformer sensor
[4] J. Fang and S. Xu, “Effects of eddy current in electrical connection in magnetic bearings,” Chinese Conference of Theoretical and Applied
surface of laminated cores on high-speed PM motor supported by active Mechanics, 2005. (in Chinese)
magnetic bearings,” IEEE Trans. Magn., vol.51, no.11, Nov. 2015, Art. [27] C. Yang and L. Xu, “Research on differential transformer displacement
ID 8207604. sensor in magnetic bearings system,” Journal of Transducer Technology,
[5] J. Tang, K. Wang, and B. Xiang, “Stable control of high-speed rotor 2005. (in Chinese)
suspended by superconducting magnetic bearings and active magnetic
bearings,” IEEE Trans. Ind. Electron., vol. 64, no. 4, pp. 3319-3328, Apr.
2017.
[6] J. Sun, Z. Ju, C. Peng, Y. Le and H. Ren, “A novel 4-DOF hybrid
magnetic bearing for DGMSCMG,” IEEE Trans. Ind. Electron., vol. 64,
no. 3, pp. 2196-2204, Mar. 2017.
[7] B. Han, S. Zheng, H. Li and Q. Liu, “Weight-reduction design based on
integrated radial-axial magnetic bearing of a large-scale MSCMG for
space station application,” IEEE Trans. Ind. Electron., vol. 64, no. 3, pp.
2205-2214, Mar. 2017.
[8] J. Sun, C. Wang and Y. Le, “Designing and experimental verification of
the axial hybrid magnetic bearing to stabilization of a magnetically
suspended inertially stabilized platform,” IEEE/ASME Trans.
Mechatronics, vol. 21, no. 6, pp. 2881-2891, Dec. 2016.
[9] Y. Maruyama, T. Mizuno, M. Takasaki, Y. Ishino, T. Ishigami and H.
Kameno, “Estimation method and measurement bandwidth of gyroscopic
sensor using active magnetic bearing,” in Proc. 33rd Annu. Conf. IEEE
Ind. Electron. Soc., Taipei, Taiwan, 2007, pp. 2230-2235.
[10] M. Zong, L. Ba and F. Wang, “Study on novel measuring method for rotor
axial displacement with displacement sensor fixed on radial direction in
magnetic bearing,” Proc. of the 11th Int. Conf. on Electrical Machines
and Systems Volume, 2007, pp. 726-729.
[11] V. Chaturvedi, M. R. Nabavi, J. Vogel and S. Nihtianvo, “Demodulation
techniques for self-oscillating eddy-current displacement sensor
interfaces: A review,” IEEE Sensor J., DOI 10.1109/JSEN. 2017.
2677526.
[12] X. Li, B. Gao, W. L. Woo, G. Y. Tian, X. Qiu and L. Gu, “Quantitative
surface crack evaluation based on eddy current pulsed thermography,”
IEEE Sensor J., vol. 17, no. 2, pp. 412-421, Jan. 2017.
[13] C. Ye, Y. Huang, L. Udpa, S. Udpa and A. Tamburrino,
“Magnetoresistive sensor with magnetic balance measurement for
inspection of defects under magnetically permeable fasteners,” IEEE
Sensor J., vol. 16, no. 8, pp. 2331-2338, Apr. 2016.
[14] Q. Tang, D. Peng, L. Wu and X. Chen, “An inductive angular
displacement sensor based on planar coil and contrate rotor,” IEEE
Sensor J., vol. 15, no. 7, pp. 3947-3954, Jul. 2015.
[15] H. Xiong and Y. Dong, “Non-contact measurement of electrolyte solution
with a passive coil placed in inductive coupling system,” IEEE Sensor J.,
vol. 16, no. 11, pp. 4238-4245, Jun. 2016.
[16] C. Franco, J. Acero, R. Alonso, C. Sagüés and D. Paesa, “Inductive sensor
for temperature measurement in induction heating applications,” IEEE
Sensor J., vol. 12, no. 5, pp. 996-1003, May. 2012.
[17] A. Danisi, A. Masi, R. Losito and Y. Perriard, “Modeling of
high-frequency electromagnetic effects on ironless inductive position
sensor,” IEEE Sensor J., vol. 13, no. 12, pp. 4663-4670, Dec. 2013.
[18] A. Masi, A. Dasini, R. Losito and Y. Perriard, “Characterization of
magnetic immunity of an ironless inductive position sensor,” IEEE
Sensor J., vol. 13, no. 3, pp. 941-948, Mar. 2013.
[19] S. Fericean and R. Droxler, “New noncontacting inductive analog
proximity and inductive linear displacement sensors for industrial
automation,” IEEE Sensor J., vol. 7, no. 11, pp. 1538-1545, Nov. 2007.
[20] H. Sha, “Design and experiment study of a micro active magnetic
bearing-rotor system,” Tsinghua University, 2006. (in Chinese)
[21] Z. Duan, Z. Wang and Z. Hu, “Application and research of AD698 in the
DGC-6PG/A differential inductance displacement sensor,” Modern
Electronics Technique, vol. 31, no. 4, Feb. 2008. (in Chinese)
[22] X. Wang, “Research on inductance displacement sensor of magnetic
bearings,” Wuhan University of Technology, 2009. (in Chinese)
[23] S. C. Chen, D. K. Le and V. S. Nguyen, “Inductive displacement sensors
with a notch filter for an active magnetic bearing system,” Sensors, vol.
14, no. 7, pp. 12640-12657, Jul. 2014.
[24] M. Recheis, J. Nicolics, H. Wegleiter, B. Schweighofer and P. Fulmek,
“Evaluation of inductive displacement sensors for a basic active magnetic

1530-437X (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.