Analysis and control of computer cooling fan noise

Abstract

The problem of noise radiation from large turbo machines in aerospace applications has received a lot of attention, but the same is not true for small ventilation fans. The noise from the small fans shares many features with that of the large turbomachines, but there are important differences. For example, Gutin noise can be ignored for the small fan, but it may be significant for the turbo-fan of aircraft engine. The other reason for the difference is that technical measures allowed for large fans may not be feasible for small fans due to the cost. The main objective of this study is to find feasible techniques to reduce small fan noise through a thorough understanding of the noise mechanisms which are previously lacking. This thesis is divided into three parts: the study of the source mechanisms and their separation, passive noise control, and active noise control. The mechanisms of noise radiated by a typical computer cooling fan is investigated both theoretically and experimentally focusing on the dominant rotor-stator interaction. The unsteady force generated by the aerodynamic interaction between the rotor blades and struts is phase locked with the blade rotation and radiates tonal noise. Experimentally, synchronous averaging with the rotation signal extracts the tones made by the deterministic part of the rotor-strut interaction mechanism. This averaged signal is called the rotary noise. The difference between the overall noise and rotary noise is defined as random noise which is broadband in the spectrum. The deterministic tonal peaks are certainly more annoying than the broadband, so the suppression of the tones is the focus of this study. Based on the theoretical study of point force formulation, methods are devised to separate the noise radiated by the two components of drag and thrust forces on blades and struts. The source separation is also extended to the leading and higher order modes of the spinning pressure pattern. By using the original fan rotor and installing it in various casings, the noise sources of the original fan are decomposed into elementary sources through directivity measurements. Details of the acoustical directivity for the original fan and its various modifications are interpreted. For the sample fan, two common features account for most of the tonal noise radiated. The two features are the inlet flow distortion caused by the square fan casing, and the large strut carrying the electric wires for the motor. When the inlet bellmouth is installed and the large strut is trimmed down to size, a significant reduction of 12 dB in tonal sound power is achieved. These structural corrections constitute the passive noise control. However, the end product still features the leading mode drag noise. Further reduction of this noise is left to the active noise control. The feasibility of the active noise control technique is demonstrated for the cancellation of both thrust and drag noise radiated at their leading modes. An open loop, feed-forward system is used to maximize the simplicity of the rig in order to deliver an appropriate technology for a small ventilation fan. The leading mode configurations are constructed by re-designing the struts. The control rig consists of three components, a miniature electret microphone used as a rotation sensor, ordinary loudspeakers, and a bandpass filter with adjustable amplitude and phase delay. The miniature electret microphone measures the unsteady aerodynamic pressure on the fan casing, and it eliminates the possible acoustic feedback caused by the secondary loudspeaker. Its smooth and rich blade passing frequency content also allows the use of low-order filter. For the thrust noise rig, in which the number of rotor blades is equal to the number of struts, the sound power of the BPF tone is reduced by 14.8 dB, while the second BPF is reduced by 9.8 dB. For the drag noise rig, which is based on the improved version of the original fan, a 12.7 dB reduction in BPF tone is achieved. Note that the noise reduction for the second rig is on top of the passive noise control which has already reduced the original fan noise by 12 dB. For both rigs, the residual noise is analysed and it is found that the variation of the fan noise radiation from one rotational cycle to the next is the main reason that causes the mismatch between the antisound constructed from the signal input from the past cycle and the actual sound radiated at the present cycle. This variation is believed to be present even though the rotational speed of the fan is held absolutely constant, and the mechanism behind this could be rooted in the turbulent flow through the fan. A single loudspeaker is used for the thrust noise radiation and a mimimum of two loudspeakers are needed for the drag noise control. It is postulated that, for a general case where both drag and thrust noises are present at the leading modes, perhaps at different frequencies for each component, three loudspeakers would be needed to construct the antisound to achieve the global noise control for a small axial flow fan.