234
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 43, NO. 1, JANUARY 2015
Development of a Highly Precise Micronewton Thrust Balance Franz Georg Hey, Andreas Keller, Claus Braxmaier, Martin Tajmar, Ulrich Johann, and Dennis Weise Abstract— In this paper, we present our ongoing micronewton thrust balance development, which fulfils the laser interferometer space antenna requirements. In the context of the development of highly precise thrusters for attitude control of satellites for future space missions, test facilities for the characterization and qualification of thrusters need to be developed in parallel. The presented thrust balance has a resolution of 0.1 µN in a bandwidth between 1 and 10−3 Hz. As a general measurement principle, we chose a pendulum balance. The setup consists of two pendulums to enable a common mode rejection. To suppress the eigenfrequency of the pendulums, a damping system based on an eddy current brake is part of the balance assembly. A heterodyne laser interferometer is used as the translation sensor. Different measurements were performed to investigate the noise performance of the pendulum. The results are presented and analyzed. The measurement system was used to measure the thrust of a micro-high efficiency multistage plasma thruster. Index Terms— Electric propulsion, force measurement, LISA, micro thruster, micronewton thruster characterization, thrust balance, thruster testing.
LISA FEEP RIT HEMP-T DUT DWS FPGA DDS CLK Freq PLL FIFO ESC PSD
N OMENCLATURE Laser interferometer space antenna. Field electric emission propulsion. Radio frequency ion thruster. High efficiency multistage plasma thruster. Device under test. Differential wavefront sensing. Field programmable gate array. Direct digital synthesizer. Reference clock. Frequency. Phase lock loop. First in first out memory. Electrostatic comb. Power spectral density.
Manuscript received December 1, 2013; revised August 7, 2014 and October 23, 2014; accepted November 4, 2014. Date of current version January 6, 2015. F. G. Hey is with Airbus Defence and Space, Friedrichshafen 88039, Germany, and also with the Technische Universität Dresden, Dresden 01069, Germany (e-mail:
[email protected]). A. Keller, U. Johann, and D. Weise are with Airbus Defence and Space, Friedrichshafen 88039, Germany (e-mail:
[email protected];
[email protected];
[email protected]). C. Braxmaier is with the Center of Applied Space Technology and Microgravity, University of Bremen, Bremen 28359, Germany, and also with the German Aerospace Center, Institute of Space Systems, Bremen 28359, Germany (e-mail:
[email protected]). M. Tajmar is with the Technische Universität Dresden, Dresden 01069, Germany (e-mail:
[email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TPS.2014.2377652
I. I NTRODUCTION
O
VER the last decades, the need for highly precise attitude control of satellites became more and more important, especially due to the growing number of proposed future scientific space missions, which relay on precise attitude and position control to compensate the residual drag and electromagnetic radiation pressure. Possible mission examples are the LISA, Gaia, Euclid, or Darwin space telescope [6], [13]. In reference to this, different micronewton thrusters are under development, e.g., FEEP, micro-RIT, micro-HEMP-T, or cold gas systems [8], [10], [19]. In parallel to the thruster development, different organizations have started to develop and build a sufficient test infrastructure to characterize the micronewton thrusters [3], [11], [12], [18]. Challenging performance requirements are thrust level, accuracy, dynamic response, and thrust noise. For example, LISA requires a thrust range from 1 to 100 μN with a thrust resolution of 0.3 μN in a bandwidth from 1 to 10−3 Hz [1], [6], [19]. Aiming at the most challenging thrust noise requirements of LISA as a reference, especially at low frequencies, we identified a lack of available test facilities. Therefore, takings advantage from our heritage in the field of highly precise metrology [2], [9], we decided to build up a highly symmetrical pendulum balance with a heterodyne laser interferometer as optical readout which fulfils the LISA requirements, mentioned above, and which can be used to characterize the different types of micronewton thrusters. In this paper, we describe the development, design, and testing of a micronewton thrust balance that is capable to fulfil the LISA mission requirements. Noise measurements with deactivated thruster and the first thrust measurements of a micro-HEMP-T are presented. II. M ICRONEWTON T HRUST BALANCE S ETUP Presently, only a few micronewton thrust balances are under development or operational [11], [18]. The predominant measurement principle is that a force F generated from the device (usually a thruster) under test DUT deflects the measurement assembly by x. With a known spring rate k of the setup, F can be calculated by F = k · x.
(1)
The general measurement principle follows the same rules for different balance configurations. Common configurations are torsional balances or pendulum balances. Both concepts
0093-3813 © 2014 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.
HEY et al.: DEVELOPMENT OF A HIGHLY PRECISE MICRONEWTON THRUST BALANCE
235
Fig. 1. Sketch of the balance setup. a: damper. b: bearing. c: translation sensor. d: pendulum arm structure. e: thruster or device under test. The setup consists of two full symmetric pendulums. This enables a common mode rejection.
have specific advantages and disadvantages and it has been demonstrated that both concepts are suitable for thrust measurements [11], [14], [15], [18]. For the general balance configuration, we choose a highly symmetric double pendulum. A simple sketch of the mechanical balance setup is shown in Fig. 1. The assembly consists of two pendulums: 1) a reference pendulum and 2) a measurement pendulum. On the pendulum structure, the DUT and a dummy are mounted. The physical properties of the dummy and the DUT are equal. The bearing consists of four leaf springs per pendulum. The springs operate as frictionless pivots. Furthermore, they can also be used as power or data connection for the DUT. In reference to this, the balance spring rate k consists of the spring rate of the bearing, kbearing, and the spring rate caused by the force of inertia. Therefore, k is given by k = kbearing +
m 2 · g0 · lCG Ipendulum
(2)
where m is the pendulum mass, g0 is the gravitational constant, lCG is the distance between the bearing and the center of gravity of the pendulum, and Ipendulum is the moment of inertia of one pendulum. The pendulum mass is, depending on the DUT mass (up to 2 kg), in the range between 1 and 3 kg. The whole bearing assembly is a custom design. The support structure is made of aluminum and the springs are made of steel. The dimensions of one leaf spring (lspring, wspring, tspring) are 7.5 mm × 5 mm × 0.1 mm, which leads together with the pendulum length (lpendulum , 300 mm) to the spring rate of a single spring, which is given by kspring =
3 E · wspring · tspring 2 6 · lspring · lpendulum
.
(3)
Due to the parallel connection of the four springs, kbearing is kbearing = kspring + kspring + kspring + kspring.
(4)
The advantages of the chosen balance configuration are a variable measurement resolution, which can be achieved with the use of calibration weights. The symmetrical setup enables a common mode rejection. This is important to fulfil the long-term stability requirements. A general disadvantage of a pendulum balance is the sensitivity to seismic noise. Random
Fig. 2. Picture of the whole pendulum balance assembly, outside of the vacuum chamber. (a) Assembly in back view. (b) Assembly in side view. a: damper. b: bearing. c: translation sensor. d: pendulum structure. e: thruster. f: ESC. g: threaded grid. h: gas supply.
stimuli generate a motion of the pendulum at the first eigenfrequency. A damper is implemented to avoid this. In reference to the targeted measurement bandwidth, an eddy current brake was chosen as damper. The damper assembly consists of two permanent magnets (Nd2Fe14B) which are mounted on the pendulum support structure. The magnets have typically a remanence of 1.3 T. At the conduction surface of the damper the magnetic flux density is between 0.1 and 0.4 T, depending on the distance between magnet and surface. The aluminum plates are mounted on the pendulum arm to increase the effect of the damper. The distance between the pendulum and the permanent magnets is variable. Therefore, the damping coefficient is variable and can be adapted to different boundary conditions, for example, the sensitivity. Each pendulum has its own damper. A dielectric mirror (shown in Fig. 2 as c) is mounted on the pendulum arm as sensor for the optical readout. The laser interferometer system used allows a translation readout down to the picometer regime. Therefore, the readout had no influence on the balance performance. The balance operates in open loop to keep the system complexity as low as possible. The whole pendulum balance assembly is shown in Fig. 2(a) and (b). The pictures show the balance back and side views. The pendulum arm itself is made of aluminum. The support structure, which connects the balance with the vacuum chamber, consists of item profiles. The simple item profile construction combines a flexible handling with a stiff structure. At the top of each pendulum, the eddy current brake assembly is visible. The side view shows the thruster, the power, and the gas connection. The gas supply tube is coiled up to have a well-defined length and a minimized influence on the balance spring rate. The thruster power control cable is connected to the power supply unit via
236
Fig. 3. Flow chart of the readout and data acquisition. The system consists of a master clock (CLK), a direct digital frequency synthesizer, an optical setup, an FPGA-board, and a workstation with ×86 architecture.
a leaf spring. A threaded grid, on both sides of the pendulum arm, allows variable mounting of all required components. The laser head is a redesign of the actual Airbus DS heterodyne interferometer generation [2], [9] with a wider beam distance and a pig tailed optical fiber input. The use of two quadrant photodiodes allows a DWS. The DWS enables the readout of both pendulum angles, with nanorad resolution, as well as the usual translation measurement. In contrast to the translation measurement, the result of the DWS measurement is independent for each pendulum. The differential of the DWS results is equal to the translation measurement and allows an independent crosscheck of the translation signal. Because of the simplified implementation of the translation measurement in the data processing, the DWS signal is currently only used to evaluate the translation measurements. An FPGA board is used to acquire the measurement signal. Fig. 3 gives an overview of the data handling and acquisition. To generate the required frequencies, a DDS is used. To enhance the performance of the DDS, it is locked to a CLK. The DDS generates three frequencies: 1) frequency 1 (Freq 1); 2) frequency 2 (Freq 2) as input for the laser interferometer; and 3) a reference frequency (Ref Freq) as phase reference for the phase meter and the phase controller. The value of heterodyne frequency is 10 kHz. The acquired translation signals (measurement and reference signal) are processed in the phase meter in real time, with a sampling frequency of 200 kHz. The phase controller and the PLL are also operate at 200 kHz. A down sampling to 400 Hz is processed as last step of the data handling on the FPGA. To reduce the required FPGA resources to a minimum, a workstation with an ×86 architecture is used for more complex data operations. The data is transferred via a FIFO to the ×86 system to avoid errors from timing violations. On the ×86 system all further data operations, such as thrust calculation or data storing, are processed. The ×86 system includes an analog data acquisition board which allows the control of the DUT and other needed components.
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 43, NO. 1, JANUARY 2015
Fig. 4. Summary of the balance performance. Blue-dashed line: performance of one undamped pendulum. Purple dotted-dashed line: performance of a damped pendulum. Black dotted curve: requirement. Red solid line: performance of the full balance setup with active common mode rejection.
An ESC is used to calibrate the balance assembly. One part of the comb is shown in Fig. 2(a) and (b) as f. The ESC assembly consists of a comb pair, one of which is on the pendulum, shown in Fig. 2(a) as i, and the other mounted on the support structure. Johnson and Warne [16] presented that the calibration error is minimized with such a setup due to the symmetric design of the combs. In reference to these results, no further calibration of the ESC was performed [4], [16]. The comb configuration has a range from 0 to 3181 μN, due to the maximum output voltage of the power supply. The pendulum assembly is placed in a 600 L vacuum chamber. To evacuate the vacuum tank, two 700 L/s turbo pumps are used. With a gas load of 1 sccm the pressure is ∼5 · 10−5 mbar. The pumps are connected via a damper to the chamber to uncouple the balance from pump noise. Moreover, the assembly is placed on a damped optical desk, which shields the assembly from seismic noise. III. M EASUREMENT R ESULTS AND A NALYSIS Fig. 4 shows the balance performance in different configurations. The PSD shown, √ where the thrust (in μN) normalized to the frequency (in Hz) is plotted logarithmically, provides an overview of the noise level at specific frequencies. The dotted line (black) plot presents the LISA requirement (Chapter I), which is given by the required thrust level combined with the typical LISA allocation [6]. The dashed (blue) curve presents the first noise measurement of the thrust balance, where a single undamped pendulum versus a fixed mirror was measured. Therefore, no common mode rejection occurs. The observed eigenfrequency of the pendulum is 0.77 Hz, which is close to the estimated eigenfrequency of 0.8 Hz. The amplitude of the swinging pendulum is the dominant noise term of the measurement resolution. At lower frequencies, pink noise occurs, which is limiting the resolution. Possible reasons are thermomechanical drifts and the movement of the optical table. These movements are in a nanometer regime, which cause a permanent phase drift of the measurement signal.
HEY et al.: DEVELOPMENT OF A HIGHLY PRECISE MICRONEWTON THRUST BALANCE
237
Fig. 6. Calibration of the balance. The balance translation (green solid line) and the applied force (black dashed line) are presented versus time. Fig. 5. Result of the calibration for both pendulums. The force (in N) is plotted versus the translation (in m). The top (red) scatter plot shows the result of the calibration for the measurement pendulum and the bottom (blue) scatter plot presents the results for the reference pendulum.
The purple dotted-dashed line presents the performance of one damped pendulum versus a fixed mirror (no common mode rejection). The damper suppresses the eigenfrequency. Therefore, at higher frequencies, the noise level decreases 1.5 orders of the magnitude. At lower frequencies, pink noise dominates the PSD as shown in the dotted curve. The noise measurement of the full setup, i.e., two damped pendulums, is shown as the red solid line plot. The noise level has been decreased by one order of the magnitude. The assembly is stable enough to fulfil the requirement. At higher frequencies, the resolution is limited because of a peak at 1.76 Hz. Measurements with different sample frequencies make clear that the peak is not an aliasing effect. This is possibly due to either mechanical vibrations of the support structure or other external noise. Furthermore, it could be possible that the peak is due to asymmetries between the two pendulums, which would imply that the eigenfrequencies of the pendulums are not equal and would thus mean that the common mode rejection does not apply to the eigenfrequencies. Since we measure one pendulum versus the other pendulum, the doubled eigenfrequency should be visible in the PSD. But we observed a peak at 1.76 Hz, which is 0.22 Hz away from the doubled eigenfrequency. All in all, further investigations are required to find the reason for the 1.76-Hz peak. All measurements were performed with a thruster, a high voltage, and a gas supply connected or mounted on the balance. Thus, the balance is able to measure thrust in all noise measurements, but the DUT was deactivated to measure the properties of the balance and not of the thruster. A calibration with the ESC was performed to estimate the spring rate of the balance assembly and to investigate the differences between the two pendulums. The result is presented in Fig. 5. The force (in N) is plotted versus the translation (in m). The top (red) scatter plot shows the result of the calibration for the measurement pendulum and the bottom (blue) scatter plot presents the results for the reference pendulum. The slope of the scatter plots is equal to the spring rate. A linear interpolation was performed for each
scatter plot to estimate the spring rates. For the interpolation a least square linear fit was used; the norm of residuals for the fits is smaller than 10−5 N/m. The measurement pendulum has a spring rate of 15.7 N/m and the reference pendulum has a spring rate of 18.12 N/m. The results and Fig. 5 indicate that each pendulum has a different spring rate. This implies that the pendulums are not fully symmetric. A possible reason is that the fabrication and integration tolerances of the assembly cause differences in the weight and in the mass distribution of the pendulum. Furthermore, it is also possible that the length variation of the spring rate due to thermal fluctuations, or from the assembly causes this difference. A tuning of the spring rate with calibration weights is in progress. Fig. 6 shows a calibration run of the pendulum. The translation (in nm) of the balance (green curve and scale on the left side) is plotted versus time. In comparison with this, the applied force (in micronewton, dashed black curve) is plotted versus time, too. The force was applied via the ESC. Each step was 0.5-μN and 5-s-long. A full calibration cycle is shown. The translation of the balance followed the applied force instantly. The figure shows the repeatability and the drift of the balance. At the displayed time scale, the drift is smaller than 0.1 μN. This underlines the result of the noise measurement shown in Fig. 4. The plot demonstrates that the balance can measure in submicronewton range and fulfils the required stability. The first thrust measurement we investigated was the neutral gas thrust of a micro-HEMP-T. The result is shown in Fig. 7. On the x-axis, the time is plotted versus thrust (left scale, in micronewton) and the mass flow (right scale, in sccm). The measurement started at 0.5 sccm, with steps of 0.025-sccm and 20-s long. Every mass flow step generates a thrust of 0.42 μN. In this case, the balance shows a good reproducibility and a good stability as well. The absolute value achieved of the neutral gas thrust is close to the approximated thrust. The approximation is based on the results of the Gaia thruster characterization at Onera [5]. Further investigations have to show that our measurements are reliable. The thrust stand was also used to perform thrust measurements with an active micro-HEMP-T. Fig. 8 shows one of the performed measurements. The plot presents the measured thrust (black dashed line) and the measured anode current
238
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 43, NO. 1, JANUARY 2015
performance, which were presented at the IEPC 2011 and 2013 [7], [8], but a simple comparison of the direct and indirect measured thrust is so far not feasible due to the different measurement conditions, for example, higher background pressure in the vacuum tank. Therefore, the micronewton thrust balance will be moved to a larger tank with a higher pumpage and advanced measurement capabilities (plasma diagnostics). This will enable a detailed characterization of the thruster under test by a parallel operation of the thrust balance and the plasma diagnostic, which will lead to deeper understanding of measurement uncertainties. IV. C ONCLUSION Fig. 7. Measurement of the neutral gas flow of the micro-HEMP-T. The thrust (blue bold curve) and the mass flow (black dashed curve) are plotted versus time.
Fig. 8. Example of a micro-HEMP-T thrust measurement. Plot: measured thrust and measured anode current versus time.
(green solid line) versus time. The thruster was operating with a fixed anode voltage of 500 V. The mass flow was varied from 0.9 to 0.6 sccm. At 4250 s, the thruster has been deactivated. The decreasing of the mass flow leads to a decreasing of the anode current. Due to the fact that the anode current is equivalent to the thrust [17], the measured thrust is also decreasing. Qualitatively, the thrust follows the anode current. The variations of the anode current at 250 s and between 3600 and 4250 s are also visible in the measured thrust. Between 900 and 3800 s, a mismatch of the current and the thrust can be observed. For a detailed discussion of the reasons of the mismatches and a quantification of the error, a detailed analysis of the tested thruster have to be performed. Possible reasons are related to the variation of the thruster efficiency at the different operation points, for example, the percentage of the neutral gas thrust, variation of divergence efficiency, ion acceleration efficiency, and so on. After the deactivation of the thruster, the measured thrust is not directly zero as for the anode current. This is because of the slower decrease of the neutral gas flow, which is still producing a thrust. In general, the plot demonstrates the capability to measure a real electric thruster with the thrust balance. The present results underline the expected thruster
The micronewton thrust balance at Airbus DS is able to measure submicronewton thrust and fulfils the LISA requirements. The calibration of the balance was performed with an ESC. The √ balance assembly has a measurement resolution of 0.1 μN/ Hz in a frequency range from 10 to 2 · 10−3 Hz. This allows characterization of possible thruster candidates for LISA. First tests with an active thruster were performed. To obtain a deeper understanding of the observed effects and to identify the possible measurement limitations, further tests are in progress. In the next months, the balance will be upgraded with an electromagnetic actuator to enable a closed-loop measurement. Furthermore, the whole assembly will be moved to a new vacuum infrastructure with higher pump capacities to enable operation at higher gas flow and maintaining a representative vacuum environment. Moreover, this allows the integration of other metrology equipment next to the balance. It is planned to install a Faraday array and a retarding potential analyzer. After the completed integration in the new facility, it is planned to test the balance with different thruster types. R EFERENCES [1] P. Gath, D. Weise, and M. Ayre, “LISA: Mission design description,” Astrium, 2006. [2] M. Gohlke, T. Schuldt, D. Weise, U. Johann, A. Peters, and C. Braxmaier, “A high sensitivity heterodyne interferometer as a possible optical readout for the LISA gravitational reference sensor and its application to technology verification,” in Proc. 7th Int. LISA Symp., 2009, p. 012030. [3] T. W. Haag, “Thrust stand for pulsed plasma thrusters,” Rev. Sci. Instrum., vol. 68, no. 5, pp. 2060–2067, 1997. [4] A. J. Jamison, A. D. Ketsdever, and E. P. Muntz, “Gas dynamic calibration of a nano-Newton thrust stand,” Rev. Sci. Instrum., vol. 73, no. 10, p. 3629, 2002. [5] J. Jarrige et al., “Thrust measurements of the Gaia mission flight-model cold gas thrusters,” J. Propuls. Power, vol. 30, no. 4, pp. 934–943, 2014. [6] O. Jennrich et al., LISA—Unveiling a Hidden Universe, European Space Agency, Paris, France, 2011. [7] A. Keller et al., “Parametric study of HEMP-thruster, downscaling to μN thrust levels,” in Proc. IEPC, Oct. 2013. [8] A. Keller et al., “Feasibility of a down-scaled HEMP-thruster,” in Proc. IEPC, 2011. [9] H. Koegel et al., “Interferometric characterization and modeling of pathlength errors resulting from beamwalk across mirror surfaces in LISA,” Appl. Opt., vol. 52, no. 15, pp. 3516–3525, 2013. [10] H. Leiter, D. Bock, B. Lotz, D. Feili, and C. Edwards, “Qualification of the miniaturized ion thruster RIT-μ-perspectives, program, results and outlook,” in Proc. IEPC, 2011.
HEY et al.: DEVELOPMENT OF A HIGHLY PRECISE MICRONEWTON THRUST BALANCE
[11] J. P. Luna, C. H. Edwards, J. Gonzalez del Amo, and B. Hughes, “Development status of the ESA micro-Newton thrust balance,” in Proc. IEPC, 2011. [12] K. Marhold and M. Tajmar, “Direct thrust measurement of in-FEEP clusters,” in Proc. IEPC, 2005. [13] J. Mueller, “Thruster options for microspacecraft: A review and evaluation of existing hardware and emerging technologies,” J. Appl. Phys., 2007. [14] K. A. Polzin, T. E. Markusic, B. J. Stanojev, A. DeHoyos, and B. Spaun, “Thrust stand for electric propulsion performance evaluation,” Rev. Sci. Instrum., vol. 77, no. 10, pp. 105108-1–105108-9, 2006. [15] T. Moeller and K. A. Polzin, “Thrust stand for vertically oriented electric propulsion performance evaluation,” Rev. Sci. Instrum., vol. 81, no. 11, p. 115108, 2010. [16] N. P. Selden and A. D. Ketsdever, “Comparison of force balance calibration techniques for the nano-Newton range,” Rev. Sci. Instrum., vol. 74, no. 12, p. 5249, 2003.
239
[17] G. P. Sutton and O. Biblarz, Rocket Propulsion Elements. New York, NY, USA: Wiley, 2010. [18] S. Rocca, C. Menon, and D. Nicolini, “FEEP micro-thrust balance characterization and testing,” Meas. Sci. Technol., vol. 17, no. 4, p. 711, 2006. [19] M. Tajmar, “Overview of indium LMIS for the NASA-MMS mission and its suitability for an in-FEEP thruster on LISA,” in Proc. IEPC, 2011.
Authors’ photographs and biographies not available at the time of publication.
