Can Audio Quality be Objective? [Article]
Some people prefer to check audio quality via own ears ("subjective" approach). And they are right from point of view of music final destination - aesthetical enjoyment.
Other people learn technical distortion level of the equipment ("objective" approach). And they are right from point of view: "if music is lesser distorted then sound closer to reference original".
But there is not all so simple. Read details below.
Sound quality may have 2 definitions:
-
Audio quality is distortion and noise level (objective approach).
- Audio quality is a beauty of sound (subjective approach).
Objective is measurable (with some precision) and repeatable
Subjective is perceived and repeatable/unrepeatable.
Objective approach as audio quality estimation
Second definition is pure "subjective" (personal feelings).
Theoretically "most natural" sound is original sound of acoustical source (voice, musical instrument, other). In ideal case, original sound is passed thru the system to our ear without changes. It is maximal audio quality of the system.
Original sound also contains spatial information. But modern apparatus still don't provide it entirely. Read details here.
However, first "objective" (measurable) definition is not so simple and univocal as it might appear at first glance.
We can measure figures, but we don't know exactly how it perceived by human.
Let's look to examples:
- Noise floor have different shapes for similar energy. What is more qualitative?
- Pre-ringing with lesser energy and post-ringing with higher one. What is better for ear?
Here we are faced with "subjectivity" when we try interpret figures. And first "objective" quality definition can't exactly estimate sound quality.
Different noise shape example
Let's learn different noise shape example. Two music devices have similar total energy noise. We can add "objectivity" via normalizing to ear sensitivity curve (hearing curve) [1]. Noise shape will changed.
What sound is better objectively?
What is better subjectively?
But we can't answer objectively right now: what noise floor is better to ears? Because the floors are similar enough. And we don't know, that may be better without experimental check.
We can get close to "objective" noise floor comparison via blind audio test. Despite we compare "subjective" perception.
Sometimes difference may not be distinguished.
Blind test is kind of hifi tests, that allow to get rid of "subjectivity" partially. It don't give 100% sureness, because many subtlest details can cause result bias. But such trial is better than nothing.
Pre-ringing vs. post-ringing example
After digital filtration (in resampling, as example) output signal contains ringing - artifacts, generated by signal.
Pre-ringing rise at filter output before "parent" signal. Post-ringing - after.
Pre- and post-ringing of digital filters
Intuitively we think, that pre-ringing is worse. Because it is so "unnatural". It's really strangely that distortions rise before signal.
Minimum phase filter allow to move pre-ringing energy to post-ringing area. However phase response of the filter have some non-linearity and post-ringing energy is increased 2 times.
We know all figures again. I'd say that post-ringing of minimal phase filter "2 times worse", than linear filter's one. But we don't know that it mean to our ears.
Until audio trials, of course.
Subjective perception as audio quality estimation
Subjective perception always cause more doubts, than measurements. Because recheck it other way is sophisticated enough. Even if a listener report repeatability of results.
We can estimate aesthetical enjoyment from apparatus. We feel, that "nicer" sound is like a better sound quality.
But some kinds of distortions may have "nicer" sound. Logically, we can suggest, that distorted sound have lesser "naturalness", than undistorted. I.e. recorded musical instrument, sound "nicer" subjectively, but "naturalness" is lesser.
Subjective sound quality:
analog sources may have nicer sound than digital ones
As example, currently digital systems are most exact. They have minimal distortions. But analog sources (tape, vinyl) may be estimated as "nicer" by sound. Despite with serious issues related to mechanical and medium material matters.
Vinyl, like tape, can give nicer sounding,
despite mechanical and medium material issues
Nicer sound here is result of specific kind of distortions. As example, tape have "soft compressing" issue (specifical non-linearity), that give tape sound original coloring.
Above-mentioned analog mediums may be considered as sound enhancers (sound "coloring"). But it is not relate to definition of sound quality.
Sound engineers know some almost invisible things, that can improve perceived sound quality: slight compression, slight boost of high frequencies and other.
However, if these effects are applied too obviously, they can cause lower perceived quality.
Conclusions
-
In the audio we can't achieve "100% objectivity" in sound quality estimation.
-
We can measure features. But we can't estimate by figures its impact to ears with 100% sureness.
-
We can use blind test for "objectivization" of "subjective" perception of the impact. Blind test have many subtlest details, that can cause result bias.
-
We can't throw out subjective estimation of sound quality (see goal 3). But distinction must be drawn between:
"qualitative audio" (minimal distortions of original) and "sound enhancing" (beauty via "coloring").
Audiophile Inventory's developer,
31 July 2017