In recent years, the home and commercial appliance industries have shown an increased demand not just for efficiency and quieter operation of appliances, but also for better sound quality. Traditionally, such products have been acoustically labeled by using their sound power level, and noise and vibration laboratories around the world are equipped with application-specific testing rooms and instrumentation dedicated to the measurement of the sound power spectrum. Original Equipment Manufacturers (OEMs) in the refrigeration industry give sound power specifications to their suppliers, who must ensure that their product sound power spectra and overall A-weighted level comply with such specifications.

However, compliance of a component or sub-system to a sound power specification does not guarantee sound quality when the component is installed in the final application. This happens for two reasons:

• Sound quality is not captured by a spatial and temporal average of A-weighted noise, that is, by sound power.
  
  
• The component sound power specification is often established with no consideration of the interaction between the component and the unit as a whole.

Enlarge this picture Fig. 1. Noise at 1m in front of icemaker.

The importance of those two observations can be illustrated by a sound quality troubleshooting project involving a commercial icemaker. The manufacturer of the unit had expressed concerns over sound quality to its compressor supplier. The compressor complied with the sound power specification mandated by the icemaker manufacturer, yet the unit, when run at 60 Hz line frequency, emitted an annoying, very noticeable low-frequency moan. The icemaker manufacturer then asked the compressor supplier to diagnose and solve the problem. The ice maker in question produced ice cubes. Like in a refrigerator, the compressor forces refrigerant through the condenser and the evaporator. In the icemaker, the ice tray is attached to the evaporator and the water that is forced to flow over the tray gradually freezes due to the evaporation process. Ice gradually builds up in the well of the tray. After a specified time, the compressor cycle is switched so that hot gas is cycled back to the evaporator without condensing. The evaporator pipes and the ice tray heat up rapidly, which loosens the ice cubes, which then fall into the ice bin. Unlike in a refrigerator, the compressor is always running in an icemaker and the refrigeration cycle goes through an abrupt discontinuity when the path of the refrigerant is changed to release the ice.

To understand the reason of the complaint, the problem was decomposed into a Source-Path-Receiver model, which is standard practice in the noise and vibration field. Since sources and paths are unknown at the start of the project, the first two steps of a troubleshooting project are:

• Define and map the receiver, in this case, a microphone at 1m in front of the unit.
  
  
• Work with the product engineers to understand how the system works in order to establish an efficient testing strategy that will determine the type and location of transducers, test conditions, and so on.

Enlarge this picture Fig. 2. Sound power of compressor.

The noise of the icemaker measured in front of the unit is shown in Fig. 1. The A-weighting filter, which is widely used to compute the A-weighted sound pressure level or dB(A), reduces the level of the low frequencies and leaves unchanged the level of the higher frequency bands. When introduced several decades ago, the purpose of the A-weighting filter was to provide a simple representation of the frequency sensitivity of human hearing. Since then, psychoacoustics research has revealed that using the A-weighting filter severely underestimates the annoyance of low frequency noise. For this reason, it is often useful to display the unweighted (or linear, dBlin) frequency spectrum of the measured noise along with the corresponding A-weighted spectrum.

After listening to the recorded noise and applying digital filters to remove individual frequencies, it was concluded that the reason of the complaint was a narrow tone at 238 Hz, corresponding to the 4th harmonic of the compressor pumping frequency. All engineers agreed on the fact that the tone was very loud, especially in front of the unit. Its annoyance was also due to its frequency, which was low relative to the frequency of the rest of the noise made by the unit. This meant that the tone at 238 Hz was unmasked by other noises and this made it very noticeable. Furthermore, annoyance is in general highly correlated to low frequency noise.

The measured sound power of the compressor, shown in Fig. 2, confirmed the compressor 4th harmonic as the cause of the sound quality complaint. However, the A-weighted sound power of the compressor, at 75 dB(A), complied with the customer specification of 78 dB(A). So, there was a clear disconnect between sound power requirement and sound quality of the icemaker. This can be seen again in the spectrum of the noise in front of the unit in Fig. 1, where the contribution of the 250 Hz band to the overall level is much smaller in the A-weighted than in the linear spectrum.

The question was why there existed the amplification of the 4th compressor order in the icemaker. Diagnostic tests were run with pressure transducers in suction and discharge lines, with accelerometers on the compressor housing and feet and on the icemaker basepan, and with microphones both outside the icemaker and inside the ice bin. The tests established that the base plate supporting the compressor exhibited a considerable flexural motion at 238 Hz. During the course of several tests, it was also noticed that the level of the 238 Hz tone not only changed with the temperature and pressure conditions in the system, indicating a probable acoustic root-cause, but also with the amount of ice present in the ice bin.

A quick calculation of the first few acoustic modes of the bin, due to its dimensions and to the volume occupied by the ice, provided a couple of frequencies very close to 238 Hz, suggesting an acoustic amplification occurring in the ice bin. The picture forming at that point was that the compressor was structurally exciting the baseplate, which in turn moved as a loudspeaker and radiated noise in the ice bin below, where further amplification of the 238 Hz tone occurred due to the geometry of the bin. But this only explained the path from the compressor to the front microphone; the root-cause of the excitation at 238 Hz was still unknown.

Enlarge this picture Fig. 3. Compressor cavity mode at 238 Hz.

Additional data established the right direction. Looking at 4th order slices of the pressure at top and bottom of the compressor housing during a line frequency sweep clearly indicated that top and bottom are exactly 180 degrees apart, strongly suggesting a vertical mode shape of the compressor.

A 3D acoustic model of the compressor interior geometry and housing revealed an interior vertical/diagonal acoustic mode. (See Fig. 3). Areas of red and blue indicate, respectively, peaks and valley of amplitude, just like maxima and minima of a standing wave. The root cause of the sound quality concern was therefore found: the 4th order of the pumping frequency excites an acoustic mode (or standing wave) in the suction volume inside the housing. That also explains why the icemaker did not cause noise complaints at 50 Hz line frequency, since at this frequency the compressor runs at lower RPM and its pumping harmonics (excitation frequency) do not align with the acoustic resonance of the interior cavity.

This back-and-forth oscillation between top and bottom causes the compressor to bounce on a mainly vertical axis. Due to the weak isolation provided by the compressor grommets in this frequency range, this motion is transferred to the baseplate and then amplified throughout the unit as described.

Due to the low frequency of the concern, the only practical solutions must address the problem either at the source (that is, inside the compressor) or through the structural path (compressor mounts), given that a significant increase of acoustic isolation at 240 Hz would require an unrealistic increase in weight of the icemaker panels.

There were several possible solutions to this problem; however, not all of them might be feasible from a production or cost standpoint. Those that were tested and proved effective in greatly reducing this sound quality concern included the following:

At the source:

• Suction muffler.
  
• Tuned λ/4 resonator.

Along the structural path:

• Softer grommets.
  
• Adding damping to baseplate.

This example illustrates well the shortcomings of a sound power-based approach to product sound quality design. Sound power is a very valuable measure of airborne noise radiated off a component. However, it does not account for vibro-acoustic interactions between components and between component and application. Those interactions and their effects on the overall unit noise can only be identified by decomposing the system in a Source-Path-Receiver model, which can be applied to any system from the simplest to the more complex. Another lesson learned from this project is that, like in most such projects, a simple noise concern at the receiver may be caused by a complicated combination of sources and transmission paths, all of which must be mapped for a successful countermeasure strategy.

For more information, visit: www.soundanswers.net