Psychoacoustic Evaluation of Fan Noise Dr. Marc Schneider Team Leader R&D - Acoustics ebm-papst Mulfingen GmbH & Co.KG Carolin Feldmann, University Siegen
Outline Motivation Psychoacoustic Parameters Psychoacoustic Lab Systematic Hearing Comparisons Psychoacoustics of Fan Noise, Two Examples Psychoacoustic Metric Conclusions 2
Motivation: Is today s fan noise specification sufficient? One fan with higher sound power than an other: is this fan really more annoying, noisier? Does a spectrum really gives an indication about noise quality? 3
Motivation and Psychoacoustic Parameters Up to now: objective (physical) evaluation of fan noise Sound Quality requires the consideration of subjective characteristics of human hearing A-weighted sound pressure level merely serves for coarse classification and adverse selection often unsuitable for the characterisation of the subjective felt sound quality Evaluation of additional Psychoacoustic Parameters needed Psychoacoustic Unit Description parameter Loudness N sone The subjective felt sound intensity (intensity sensation) Sharpness S acum Relation between high- and low- frequency spectral fractions Fluctuation Strength F vacil Fluctuations in the sound signal up to approx. 20 Hz modulation frequency maximum at modulation frequency of 4 Hz, ref. signal: 1 khz, 60 db(spl), m = 1, f mod = 4 Hz 1 vacil Roughness R asper Fluctuations in the sound signal between approx. 15 and 250 Hz modulation frequency maximum at modulation frequency of 70 Hz Tonality T db(penalty) Sensation of tonal components which stick out from a broadband sound spectrum often felt to be disturbing/annoying 4
Psychoacoustic Lab noise data (time signals) Eight listening stations with headphones and touchscreens Loudspeakers with subwoofers HEAD software ArtemiS Suite and SQuare for noise analysis and hearing comparisons 5
Systematic Hearing Comparisons Two methods: 1) Paired Comparison considers just one attribute in each case offers the possibility to compare two sounds directly, commonly with regard to annoyance 2) Semantic Differential addresses multiple dimensions of the sound signal use of attributes (here: 10) to describe different aspects of a noise signal for each attribute: one can choose a level on a 7-step scale Quality Evaluation Spectral content Time structure strong weak quiet loud without tones with tones non-fluctuating fluctuating high quality low quality non-annoying annoying non-whooshing whooshing non-humming humming non-droning droning high tone low tone done with a number of test persons (example: 25) 6
Psychoacoustics of Fan Noise 52 time signals of axial fans and 58 time signals of radial fans from ebm-papst data base are investigated diameter ranges from 250 to 1250 mm and rotational speed from 450 to 4000 rpm 7 Median of relative specific loudness vs. critical-band rate z (z = 0.1 Bark) shapes are similar, but fan noise is characterized more strongly by low frequencies while pink noise is dominating in high frequency regime
Psychoacoustics of Fan Noise median curves of relative roughness R and relative fluctuation strength F averaged graphs are very similar and reveal a close match to respective pink noise spectrum roughness is pronounced between 3 and 11 Bark, maximum of fluctuating strength is narrower at approximately 3 Bark exemplary outliers show how big deviations from median curves can be in individual cases 8
Example 1: Radial fans under different inflow setups Speed [rpm] Blade Number Fan 1 2015 11 Fan 2 1993 7 Fan 3 1941 11 D = 250 mm Sound pressure level Lp Loudness N 9 Fan 2: lowest levels (Lp) for straight inflow (w/o, NWT), seems to be sensitive to lateral inflow setups Fan 1 and 3 perform very similar with varying inflow configurations (with slight advantages of fan 1) Lp max. for fan 2 with N1 and H1 do not occur in the loudness diagram in contrast to Lp a lateral inflow (N1, N1d, H1) has no general negative impact on the perceived loudness N
non-annoying Example 1: Radial fans under different inflow setups non-annoying 10 spider chart as a result of hearing comparison performed as Semantic Differential: 10 attributes queried from 25 test persons on 7-step scale (-3 to +3, positive antonym at outer side) two sounds of fan 2 (green lines) deviate somehow from the others sound of fan 3, w/o is perceived as low-quality, more tonal, louder and more annoying than the others relationship between sensation of tone pitch (where this configuration has a high value) and the subjective felt quality respectively the annoyance can be presumed annoyance increases for fan 1 and 2 with N1, opposite behavior for fan 3 probably due to reduction of high frequency spectral fractions by N1
Example 2: Radial fan Variation of speed and duty point Radial fan z = 7 (backward curved) D = 250 mm n = 1390.. 3480 rpm 4 qv 2 D n 3 11 SPL LpA and loudness N have similar shapes: both diagrams reveal a flattening of the graphs and a movement of the maximum value point from low to high values of the flow coefficient for reduced speed Roughness (R) curves comparable to the loudness (N) graphs, but with maxima at low values of for all rpm fan noise becomes rougher with decreasing volume flow rate, probably due to flow separation duty point = 0.15 appears as outstanding for fluctuation strength F beginning separation ( rotating stall )
Example 2: Radial fan Paired Comparison Paired Comparison with 8 test persons (average age: 30 years) according to the method of double-sided A/B-Matrix Radial fan (D=250mm), three operation points at maximum speed (n = 3480 rpm) Besides the original sound signal (blue) a manipulated signal (red) for every duty point is considered investigation of the influence of tonal components (linked with BPF and rotating stall) original 1 = 0.29 2 = 0.24 3 = 0.15 manipulated 12
annoyance factor Example 2: Radial fan Paired Comparison sensation of annoyance (annoyance factor from bipolar scale) for three selected operation points and respective manipulated duty points (*) plotted using statistics: median values (red line), interquartile range (blue rectangle) and 1.5-times interquartile range (Whisker, black line) 1 = 0.29 2 = 0.24 3 = 0.15 13 stimulus at the optimum point ( 2 = 0.24) is evaluated best, manipulated signals allocated to a distinct annoyance factor in two of three cases merely at 3 = 0.15 slight uncertainty occurs signals without tones (*) perceived always as less annoying necessity for consideration of psychoacoustic parameter tonality T in this example
Psychoacoustic Metric Psychoacoustic Annoyance PA for synthetic and technical sounds (acc. to Fastl and Zwicker): 2 2 2.18 PA N 5 1 (0.25 log( N5 10) ( S 1.75)) 0.4 F 0. 6 R 0.4 N5 Modified Psychoacoustic Annoyance PA* including tonality T (acc. to Feldmann, Master Thesis): PA* PA (1 T) 14 shows necessity for consideration of T to account for the annoyance perception in this example relationship between annoyance factor coming from the subjective hearing comparison and PA* calculated from objective psychoacoustic parameters becomes significantly clearer (R² = 0.87)
Conclusions Physical parameter sound level and related sound spectrum characterize fan noise incompletely psychoacoustic parameters needed to take into account subjective noise sensation Averaged psychoacoustic spectra are similar to pink noise pink noise could be used for hearing comparisons with reference sound Hearing comparisons are conducted in form of two methods: semantic differential and paired comparison Annoyance level is decisively associated with objective psychoacoustic parameter loudness, tonality and fluctuating strength (especially in fan operating points near stall) First steps in finding a psychoacoustic metric of fan noise relationship between objective psychoacoustic parameters and subjective noise sensation (e.g. annoyance) Goal for the future: 1) use of psychoacoustic parameters or combinations additionally to the physical ones 2) establishing an international standard based on existing norms referring to the psychoacoustics of fan noise 15