Measurement Noise Reduction for Study of Rocket Engine Thrust Instabilities

This paper focuses on the development and assessment of methods for improving the quality of rocket engine thrust measurements to allow for accurate determination of frequency components and their magnitude. Rocket engine testing is a time consuming and expensive process requiring precise measurements. The results of such measurements are thoroughly analyzed and thus errors such as drift or excessive noise can lead to false conclusions, necessitating a careful approach. Arguably the most crucial part of an engine test is determination of the thrust curve, most often performed via resistive sensors (load cells), which are simple to use but exhibit numerous sources of errors, mostly stemming from different types of noise. Typical applications of load cells only partially address this issue, providing temperature compensation and limited noise reduction with the use of low-pass filters. Such approaches do not provide the most accuracy due to the necessary compromise between noise level and measurable bandwidth, resulting in rejection of high-frequency components, and the dominance of 1/f noise, which distorts measurements even at low frequencies. Since rocket engines can exhibit instabilities in a wide frequency range, the knowledge of which can help better understand the combustion process, elimination of these errors can be imperative for proper analysis. To this end various design rules are suggested for improving accuracy, with significant focus put on lock-in amplification via AC excitation of the load cell. Mathematical models are presented as the bases for suggested design rules and possible error sources, allowing flexible further development to adjust them to different systems. SPICE simulations are demonstrated, providing insight into the working of the proposed circuitry and allowing for easier analysis of the system, reducing the number of necessary test cases. The results of SPICE simulations of AC and DC excitation of load cells show promising results for both low- and high-frequency signals, with greatly reduced noise, allowing to accurately measure component frequencies. Additionally, the two methods have been compared experimentally on a dedicated test stand, with results similar to the mentioned simulations. These conclusions will be further validated in the next phase of the project, when a prototype will be tested in measurement of thrusts of hybrid and solid rocket motors, the results of which might allow for better understanding of the phenomenon related to thrust instabilities, effectively allowing to improve the designs. The approach can be additionally adapted to any strain gauge or various other analog measurements where high accuracy is required in a wide frequency range.