I have not read through this complete thread, but as you want my comment on loudness compensation, here it is. Apologies if I misconstrue any perspectives being discussed.
First, I noted reference to the Harman room curve "target". As I have stated several times in this forum, it is not a target, it is a result. It is not a definitive statement of sound quality, and there will be different results for different loudspeakers of comparable sound quality, mostly depending on their frequency-dependent directivity. Only below the transition frequency does the room curve become reliable information. See Section 12.3.5 in the 4th edition of my book.
Second, equal loudness curves are not "engineering data", they are statistical data pertaining to auditory threshold and loudness-balancing experiments conducted in anechoic chambers using many listeners - the variations are significant among the listeners, so where do "you" and "I" fit in. The loudness balances were made to 1 kHz pure tones, so the relationship to broadband, complex and ever-changing music is not a linear one. They are what they are, and the only real information they convey is crude generalizations about spectral balance as a function of sound level.
Fletcher and Munson used poorly calibrated headphones, so their data are outliers. Section 9.4 in the 4th edition of my book describes it. It is also in earlier editions. What is not in all editions is the following set of curves that I created in 1973 to try to explain why loudness controls/compensation cannot work for any program that has dynamic range. I will quote from the 4th edition:
"For a sound reproducing system to accurately portray what was heard in a live performance it must have a uniformly
flat frequency response and reproduce the sounds at or close to the original sound levels, so that our built-in loudness processes operate on the sound in the same way. This is where complications set in, because that is rarely the case. Even if our sound systems have flat frequency responses, the playback sound levels are not likely to be what they were at live performances or in recording control rooms where important artistic decisions were made. For recreational listening, sound levels are almost always lower, often much lower, than the “original” sound.
To understand this issue it is necessary not to pay attention to the shapes of the curves, but instead to the
differences in their shapes at different sound levels. Two things happen as playback sound levels are reduced:
- The bass frequencies be disproportionately reduced in apparent loudness compared to the rest of the spectrum, and
- Progressively more very low frequencies will fall below the hearing threshold, becoming completely inaudible."
Figure 9.6 The frequency-response changes needed to maintain the apparent spectral balance and audibility of sounds having flat spectra at original sound pressure levels of 40, 60, 80 and 100 dB SPL when the loudness level is reduced by (a) 10 phons, (b) 20 phons and (c) 30 phons. Derived from Robinson & Dadson (1957) equal-loudness contours. Figure 4.6(d) shows compromise curves drawn between the 60 and 80 dB predictions. Data shown in (a) to (d) are from Toole (1973).
Figure 9.6 shows results for three loudness reductions. A 10 phon reduction from a high “reference” level (a) is a substantial change (approximately half loudness), but one that is not at all uncommon. Most people would still consider this to be “foreground” listening. A 20 phon reduction (b) yields “background” music, and a 30 phon reduction (c) is for “ambiance”. It is immediately obvious that no single loudness compensation curve can work for music that has significant dynamic range. The low frequency boosts are substantial to maintain audibility of low-level sounds and these would render high level sounds grossly bass heavy. Clearly condition (a) will not be satisfied, but it may be worth considering a compromise solution, and some musical realities may moderate what is needed. First, low bass sounds are most often in high-level musical passages because even in live performances the sound levels must be high in order for these sounds to be audible – Figure 9.5(a). It therefore seems reasonable to focus the compromise compensation on the higher-level components, such as the 60 and 80 dB SPL portions.
Figure 9.6(d) shows the result of plotting compromise curves between the 60 and 80 dB curves in each of (a), (b) and (c). These look quite practical. It is important to note that no great changes are required at high frequencies. Many loudness compensation devices over the years have significantly boosted both bass and treble as sound level was reduced. With such controls, sound quality is diminished – a result of misinterpreting the equal-loudness contours. With 10 or 20 phon reductions essentially no high-frequency compensation is needed.
This esoteric discussion might be interesting, but it all is an approximation based on an approximation (the obsolete Robinson and Dadson contours), based on experiments done with pure tones, not broadband music. If there is a message, it is that attempts at loudness compensation for playback at less than original sound levels are destined to be imperfect, but they might contribute to improved entertainment simply by keeping some low bass sounds from being less audible or disappearing altogether."
