New Methods for Quantifying Sonic Performance

Part One: Improved Methods for Estimating Sonic Differences in Audio Systems

Equipment report
Solid-state power amplifiers,
Tubed power amplifiers,
Solid-state preamplifiers,
Tubed preamplifiers,
Digital-to-analog converters,
New Methods for Quantifying Sonic Performance

Relationship Between Height Reproduction and Up-and Down-Sampling
During the course of our investigation of the effect of up-sampling on SQ, we noticed that certain instruments (for example, the harp in Chabrier’s España, track 1, specifically the harp section from 39 to 45 sec; Acousence ACO-DF 41610) or the ranks of singers, especially the sopranos, in Misa Criolla (track 5, specifically the upper rank of sopranos occurring from 3 min, 12 sec to 3 min, 23 sec; LIM K2HD 040, UD version) seemed to vary in height in accordance with our subjective quantitative scale of SQ. It occurred to us that height alone might represent one of those characteristics that accurately reflect overall recorded quality. Therefore, we repeated our comparative listening tests, but this time attempted to estimate the height of these sonic markers in our standard test tracks. To make this easier, we created a physical scale marked off in 1/2" increments that hung from the ceiling next to our speakers. The baseline zero height was set to the center of the midrange driver of the B&W 802s. In this way we could visually estimate height in inches of these instruments or voices for the various sampling frequencies. When the height results were plotted among all of these lines suggests that we are indeed measuring the same degree of SQ using height as when we perform subjective, multifactorial judgments. These results also validate the accuracy of the sonic criteria used and our capabilities in making subjective judgments. With enough repetitions of both height and subjective SQ assessment conducted on separate days, one can establish mean and standard errors for both methods, and hence reproducibility. Assuming this data consistently shows parallelism, one can extrapolate between the two semi-log plots from height measurements to subjective SQ points. This saves time and controls for day-to-day physiological/neurological variations in subjective judgments.

Height measurement has proven to be a far easier, quicker, and more objective method to estimate audio quality than other techniques we have used. As far as we know, this is a totally new and quantifiable method for assessing SQ and has the advantage of being more consistent over longer periods of time. The benefit of this method is that it allows the listener to make comparable judgments from one listening session to the next even when such listening sessions are days apart. In addition, this method is more resistant to mental fatigue and confusion when performing extended A/B comparisons.

Naturally, there could be some room limitations to this height method of estimating SQ. For example, in System 1 the height of the drop ceiling is 7'. In the case of the Chabrier (harp) and Misa Criolla (sopranos in chorus) test tracks, once the height exceeds the interface between the wall and ceiling, the height imaging folds forward along the ceiling towards the listener. Under the conditions of these experiments the sopranos can fold forward by as much as 36 inches. In order to ascertain what the maximum height might actually be, we tested the Misa Criolla track on a system using B&W Signature Diamonds in a high-resolution system (judged to be similar to the sound in System 1) located in a great room with a two story high ceiling. The maximum height of the chorus observed in this room was approximately 100" above the floor and about 70" above the center of the semi-logarithmically as we did for the results in Figure 2, we also obtained a linear relationship, with very similar slopes. Representative examples of such results are shown in Figure 3.

It is noteworthy that all four slopes derived from the Acousence recording are quite similar to each other as well as remarkably similar to those shown in Figure 2B. The degree of similarity Signature Diamond midrange driver. In System 1, the maximum height of the choir calculated as the ceiling height plus the distance folded forward along the ceiling was 120" and about 72" above the center of the 802 Diamond midrange driver. It is interesting the degree to which these two measures of height agree when normalized to the midrange height for each speaker. We thought at the time that this might be mere coincidence. However, in recent experiments (to be reported in a future article) it appears that adding the distance to the ceiling plus the distance forward into the room along the ceiling does give a valid indication of the true maximum height of the highest choir members, the sopranos. So far, with even more tweaks to our systems, we have observed a maximum forward projection of 82 inches. (We have recently commissioned experimental recordings designed to enable us to make real-world-accurate height measurements. These recordings have taught us that microphone placement can distort actual height reproduction; thus we cannot know the true vertical location of the sopranos in Misa Criolla even if they were on typical risers behind the soloist.) The ability to portray height is not confined to the Bowers and Wilkins speakers. We found it can be heard with different combinations of speaker types and listening rooms, such as in System 2 using Paradigm speakers, in a fourth system using Quad 2905 electrostatic speakers, and two additional locales with Infinity IRS V speakers and Roger Sanders electrostatic speakers in which vertical height was not room limited.

Ultimately, the reality of the performance itself and the manner in which microphones accurately capture the vertical dimension dictates the valid upper limit of height estimates. Therefore, if one wishes to use height measurements to compare equipment or system tweaks, the performance of the system must be configured so that the range of differences falls within the height limits of the recording itself. This problem can be solved by sonically handicapping the system. For example a lower-quality power cord could be inserted or one set of support footers could be removed. Under these conditions you can then use height to measure the variable under test. The fifth track in the Misa Criolla recording turns out to have captured the greatest height information we have yet found. For this reason we have gradually shifted from using the harp in the Acousence track to the sopranos in Misa Criolla and this will be reflected in the measurement applications in Part 2 of this three-part series.

There is at least one additional variable that must be considered. We ran these height tests with seven different experienced listeners. While all listeners could discern height differences, there were significant differences in the absolute maximum height estimates among them. Two obvious variables that can affect audible perception are pinna shape and ear canal diameter. In the latter case, it is known that the diameter of the auditory canal determines its resonant frequency. Exactly which frequencies are emphasized (or amplified) may well determine in a given individual the hearing acuity of different midrange frequencies, thus affecting sensitivity to subtle midrange differences. So clearly we must remain cautious in predicting whether everyone can replicate our results to the same quantitative extent. In other words, depending on system and hearing acuity, your mileage may vary.

In Part 2 in Issue 248 we will continue exploration of our sonic measurement system as we apply this methodology to demonstrate the magnitude of three different influences on system SQ. Then in Part 3, we apply both our objective and subjective techniques to resolve the question of Mac versus PC SQ.

Featured Articles