ACTUATOR 2024 ∙ 13. – 14.06.

2024 ∙ Wiesbaden

Improving dynamic accuracy of closed loop bandwidth of piezo mech-

anisms with advanced control laws and embedded eddy current posi-
tion sensors
Clément Cote, Alexandre Pagès, Olivier Sosnicki, Mathieu Castruccio, Timoteo Payre, Guillaume Mansuy, Xavier De
Lepine, Cedrat Technologies, Meylan, France.
[email protected]

Abstract
Fast Steering Mirrors (FSM) play an important role in diverse optical applications for beam stabilization, image sharpen-
ing or dynamic beam steering. These applications require the FSM to have a precise response across a large frequency
range, thus the need of a dynamic controller. But the controller is ultimately limited, among other factors, by the perfor-
mance of the sensor: its ability to measure accurately the required physical value and with a sufficient bandwidth. The
compact, dynamic and precise embedded Eddy Current Sensors (ECS) developed by Cedrat Technologies, combined with
a model-based controller allow a two and a half times improvement on the bandwidth compared to current sensing tech-
nology, while being more precise overall.

1 Introduction
The need to increase the dynamic accuracy of Fast Steering
Mirrors (FSM) is limited by multiple factors : the mechan-
ical resonance frequency of the mechanism, the observa-
bility of the sensor or the performance of the control loop.
CEDRAT TECHNOLOGIES (CTEC) usually embeds
Strain Gauges (SG) in its actuators and mechanism. SG
measure the strain of actuating piezoelectric ceramics,
which is an indirect estimation of the mechanism position. Figure 1 TT60SM-ECS and its CAD model
SG sensors have the advantages of being compact, accurate
and easily integrable. Nevertheless, as SG give an in-direct
measure, their dynamic response is somewhat different
from the one of the mechanism positions, especially around
and above the first resonance frequency of the mechanism.
This lack of observability can be detrimental to the perfor-
mance or the dynamic precision of the controlled mecha-
nism.
CTEC has developed and embedded new position sensors
to overcome these difficulties. A prototype of its tip-tilt
mechanism TT60SM is integrated with Eddy Current Sen-
sors (ECS) to measure the position of its mirror. To fully
use the improved observability given by the new sensors, Figure 2 PCBA of the embedded ECS
it is possible to use full state feedback controllers to im-
prove the bandwidth and dynamic precision of the closed-
loop controlled system.
2.2 Driver
This mechanism can be driven by the CCBu40, a
standalone driver for piezoelectric mechanism, that can
2 System description supply up to 500 mA between -20 and +150 V. With this
configuration, a dynamic up to a few hundreds of Hertz can
2.1 Mechanism be reached.
The TT60SM is a standard Tip-Tilt mechanism of
CEDRAT TECHNOLOGIES (CTEC), that has a stroke of 2.3 Sensor conditioning
20 mrad on its axis. This prototype TT60SM is equipped with embedded Eddy
Current Sensors (ECS) in addition to the SG sensors.
The SG sensors are conditioned by the embedded condi-
tioner of the CCBu40.

203
 ACTUATOR 2024 ∙ 13. – 14.06.2024 ∙ Wiesbaden

These sensors are conditioned by a dedicated custom-made

electronic prototype, the ECSF. It consists of a self-oscil-
lating coil probe that generates a signal whose frequency
varies based on the inductance value of the probe coil.
This technology offers several key advantages. Firstly, it
enhances system observability, allowing for better moni-
toring and control. Secondly, it provides adjustable na-
nometer resolution, although this may come at the cost of
reduced bandwidth. Moreover, it offers redundancy capa-
bilities in degraded mode, which depend on the system's
topology. Finally, as a result of its digital design, this tech- Figure 5 Estimated frequency response of the system
nology is capable of operating at a low voltage (5Vdc), en-
hancing its efficiency and compatibility with a wide range The main resonance frequency of the system is at 375 Hz.
of power systems. Around the resonance frequency, it is possible to see the
However, the ECSF prototype comes with a significant lack of observability of the SG compared to the mirror po-
drawback: its susceptibility to cable movement between sition. The ECS are way closer to the real position of the
the probe and the conditioner. To address this issue, CTEC mirror, up to 900 Hz.
is developing a new sensor prototype. This updated design A zero-pole-gain model is fitted on this experimental data,
consists of two primary components: the probe segment in- and is used to synthesize and tune the advanced controller.
tegrates a digital front-end to isolate the measurement part,
while the conditioning segment manages data acquisition, 4 Closed loop performance and ro-
linearization, and user interface functionality for the sen-
sor. bustness evaluation
The main closed loop performance criterion is the -3 dB
bandwidth, which corresponds to the frequency at which
the gain of the closed loop transfer function goes below -3
dB.
Figure 3 New ECS sensor architecture
Robustness is characterised according to [1].
2.4 Test setup Phase and gain margin are two well known ways to char-
The rest of the test setup is composed of a Software-In-the- acterize robustness. They represent the minimal phase var-
Loop (SIL) Speedgoat platform, used to acquire data and iation (respectively gain variation) that destabilise the sys-
compute the control law in real time, and a high speed au- tem. Good stability margins are usually above 40° and 3
tocollimator to get the true mirror position of the mecha- dB respectively.
nism. The infinity norm of the sensitivity function is another ap-
proach of robustness. With L the open loop transfer func-
tion, we know that the closed loop becomes unstable as
trace of L in the complex plane goes from one side to the
other of the critical point −1 + 0 ⋅ 𝑗 (with 𝑗 the complex
variable such as 𝑗 2 = −1). A good way of defining the ro-
bustness is then to characterise the minimum distance be-
tween the open loop transfer function L and the critical
point. This distance is called the modulus margin, and its
value is min|1 + 𝐿|. With the sensitivity function defined
ω
Figure 4 Test setup schematic as S=1/(1+L), this point becomes:
‖𝑆‖∞ = 𝑚𝑎𝑥|1/(1 + 𝐿)|
Thus, to assure a robust control, one does not want the
3 System characterisation modulus margin to go below a certain value, corresponding
to a maximum value for ‖𝑆‖∞ . Usual values are that ‖𝑆‖∞
First, the system needs to be characterized in open loop,
should be below 2 to 3. This criterion has the advantage of
with the objective to obtain a model of the whole system.
taking into account variation of gain and phase of the sys-
The frequency response of the system has been estimated
tem at the simultaneously, thus better representing the ro-
using a white noise as the input signal.
bustness of the controlled system.

5 Standard achievable bandwidth

CTEC usually implement PID + filters controllers in its
electronics. This PID controller uses the SG sensor to close

204
 ACTUATOR 2024 ∙ 13. – 14.06.2024 ∙ Wiesbaden

the loop. The first controller is implemented to get a base- the SG sensor with 55° of phase margin, 10 dB of gain mar-
line of current CTEC abilities. gin and ‖𝑆‖∞ = 1.7.

5.1 Controller architecture An error between the measured value at the SG and the real
position of the mirror is visible on the Bode diagram.

6 Advanced controller
To further improve the bandwidth and to take advantage of
Figure 6 Standard control loop architecture the increased observability given by the ECS, a advanced
model based controller has been implemented.
This architecture is the standard architecture implemented
in CCBu20 or CCBu40 electronics, enhanced with an ad- 6.1 Controller architecture
ditional low pass filter to further improve stability margins
of the closed loop.
The notch and low pass filters frequency are tuned in re-
gard to the system identification.
The PID parameters are tuned with a trial and error method
on the real system. Figure 10 Advanced control loop architecture

5.2 Controller performance This controller has a full state feedback with a state ob-
server, and an integrator to reduce the static error to zero.
Two additional low pass filters at 500 Hz are in the loop,
to mitigate the effect of higher resonance frequency.

6.2 Controller performance

Figure 7 Standard control performance

Figure 11 Advanced control performance

This controller is able to achieve a bandwidth of 448 Hz on

the ECS, above the first resonance frequency at 375 Hz.

Figure 8 Standard control open loop frequency response

Figure 12 Advanced control open loop frequency re-

sponse

Figure 9 Standard control sensitivity function

With this architecture, it is possible to achieve 179 Hz of

bandwidth (about half of the first resonance frequency) at

205
 ACTUATOR 2024 ∙ 13. – 14.06.2024 ∙ Wiesbaden

8 Literature
[1] S. Skogestad et I. Postlethwaite, Multivariable feed-
back control: analysis and design, 2. ed., Repr. Chich-
ester: Wiley, 2010.

Figure 13 Advanced control sensitivity function

This performance is achieved with stability margins of 7.7

dB for the gain and 53° for the phase. ‖𝑆‖∞ = 1.7, so the
increased bandwidth is achieved with the same robustness
margin as the standard controller.

Here the response measured by the ECS is a better image

of the real displacement of the mirror. The better observa-
bility given by the ECS allow the control at higher band-
widths, after the first resonance frequency. Moreover, the
advanced control loop provides active damping, which is
not possible with the standard control architecture, as the
controller as a lower bandwidth and the notch filter pre-
vents any action at the resonance frequency.

7 Conclusion
The use of ECS on the TT60SM improves the controllabil-
ity of the mechanism on several aspects.
This sensor gives a better representation of the mirror dis-
placement compared to the usual SG sensor.
This improved measurement allows the efficient use of ad-
vanced model-based control law, which can control the
mirror position accurately above the resonance frequency
of the mechanism.
This results in a two and a half times improvement of the
bandwidth without significative reduction in robustness
margins.

SG + PID ECS + Ad-

vanced con-
trol
Bandwidth 179 Hz 448 Hz
0.5 x fr 1.2 x fr
Gain margin 10 dB 7.7 dB
Phase margin 55° 53°
‖𝑆‖∞ 1.7 1.7

The authors are grateful for the support of

the French government that fund part of the
work presented in this paper within the
frame of CLEOPATRE project and
FRANCE 2030.

206