Page 65 - Spring2022

(0-39%). Frequency response measurements of the head- phones were made using an ear simulator according to IEC 60318-1 (2009) equipped with a custom pinna that better simulates headphone leakage on humans.
In each category, the average frequency response for the headphones is plotted with the Harman target curve and the error curve that is the difference between the two (see Figure 3). Figure 3, black dotted line, is a regression line that best fits the error curve. The relationship between objective and subjective headphone measurements seems clear: the more the frequency response of a headphone deviates from the Harman target curve, the lower the lis- teners rated its sound quality.
Alinearmodelwasdevelopedthatpredictsheadphonepref- erence ratings using two variables based on the standard deviation and the absolute slope of the error curve. The correlation between the predicted and measured ratings is 0.86, with an error of 6.7 ratings on a 100-point scale. A similar model was developed for IE headphones that pro- duces slightly better (r = 0.91) predictions (Olive et al., 2016).
The two models were used to predict preference rat- ings for 158 headphones, including AE, OE, and IE types (Olive et al., 2018b). Figure 4, left, shows the aver- age magnitude response, standard deviation, Harman target, and error response curve for each headphone type. Figure 4, right, plots the retail price versus the predicted preference rating for each headphone model tested. On average, the AE headphones come closest to the Harman target and produce the highest preference ratings, the OE headphones are the worst, and the IE headphones fall in between. The retail price of a headphone is not a good indicator of its sound quality based on the relatively low correlation values shown here.
These findings are generally in agreement with those reported by Breebaart (2017). The two studies together provide evidence that headphone designers are aiming at a target curve that is closer to the Harman target than the DF or FF target curves recommended by the cur- rent headphone standards. Figure 5 shows the average response of the 82 AE headphones in Figure 4 compared with the Harman, DF, and FF target curves (Møller et al., 1995). Although the DF and FF targets specify a flat response below 200 Hz, the average AE headphone and Harman targets have 5-6 dB more bass, which better
approximates the preferred in-room response of a full- range loudspeaker calibrated in a typical listening room.
Conclusions
Our understanding of the perception and measurement of headphone sound quality has not kept pace with consumer demand and expectations. Two independent studies measured over 400 headphones and came to similar conclusions: there is little correlation between the price of a headphone and its frequency response, the single best indicator of its sound quality. Most profes- sional and consumer headphone designs today do not comply with the FF and DF targets recommended by cur- rent headphone standards, which warns “the objective methods whose results bear good relation to those from subjective assessments are under research stage” (see IEC 60268-7, 2010, Section 8.6.1). The research stage is largely completed, the results are in, and the headphone standards need to be updated.
Listeners largely agree on what makes a headphone sound good. For stereo reproduction, the preferred headphone target approximates the in-room response of an accurate loudspeaker calibrated in a semireflective room. This makesperfectsensebecausestereorecordingsareintended to sound best through accurate loudspeakers in semireflec- tive rooms. What makes a headphone sound good is the same as what makes a loudspeaker sound good.
Figure 5. Proposed headphone target curves normalized at 500 Hz: the Harman AE headphone target (green), two diffuse-field (DF; orange and black) and free-field (FF; dashed) targets (Møller et al., 1995), and the average frequency response (blue) of 82 different models of AE headphones (Olive et al., 2018b).
Spring 2022 • Acoustics Today 65