MUSINGS: A quick discussion on what's OBJECTIVE and what's SUBJECTIVE (alas, more MQA-related stuff).

Infographic, from pediaa.com.

Hey guys, very busy this past week and about to go out of town... Nonetheless, it has been fun participating in discussion forums this week. Therefore, I thought I'd just fire out a "quick" post this morning.

I see some rather disturbing back-and-forth arguments on the recent Stereophile post on MQA from folks like dalethorn and boulderskies that IMO reflects a fundamental disagreement on the differences between what the words "subjective" and "objective" means when we use them to address ways of understanding audio (among other ways of discerning knowledge in this world). Obviously, unless we are all on the same page with definitions, we won't be going anywhere with debates and arguments!

At first, I was going to just make this comment a forum post, but given the length and for future reference, I'll just put it here instead of a typical comment buried deep in a thread. We can potentially get deeply philosophical about this, but let's see if we can just hit the main points at least...

First, let's get some dictionary definitions out there (as per Google):

OBJECTIVE:
(of a person or their judgment) not influenced by personal feelings or opinions in considering and representing facts.

SUBJECTIVE:
based on or influenced by personal feelings, tastes, or opinions.

Boulderskies bluntly said in one of the comments:

Objective means graphs & charts; Subjective means, "Does it sound good?"

Clearly it's not so simple. When someone shows graphs and charts based on what a machine was able to measure, we can say at some level that this is "objective" because the data is likely a quantitative result that originated apart from direct human input. Not all charts and graphs are objective though because they could just be a reflection of opinion (eg. graphing a bunch of subjective scores like "4/5" for "recording quality" comparing album reviews). Of course not all machine-based results are accurate based on whether the test is appropriate and the operator had performed the test soundly - bias can creep into measurements. Furthermore, measurements can show variability, and issues like inter-test reliability may need to be considered as well. Rigorous set-up, using a measurement device of adequate resolution, and appropriate use of test signals must be followed in order to ensure objective tests reflect a non-biased result.

In the case of MQA, with objective testing, using known signals, we can detect subtle distortions and show imaging artifacts with MQA filtering (as reviewed recently). Likewise, we can detect the limitation of resolution with MQA by using actual MQA encoded/decoded music and measuring the bit-depth reduction. These things are facts, replicated by others and not just my opinion. (In "fact", this morning, I received some data from others having done the null measurements showing bit-depth loss with MQA from about 10 albums.)

From the facts, we can deduce some inner workings of a system and form opinions around quantitative and likely qualitative implications since we also have knowledge of human auditory perceptual abilities we can try to map the findings to. Furthermore, we can figure out for ourselves how MQA works independent of what MQA (or Bob Stuart) tells us because we can use our knowledge of audio and computer science fundamentals to understand the mechanisms being employed, although not down to every proprietary detail.

Dale's PDF notes on MQA and the comments by boulderskies invoke subjective "analysis". They pertain to expressions of personal tastes and opinions either when listening individually or desiring to know what others opine.

We all have opinions. But what can be deduced from opinions is a different level of "knowledge" than objective analysis. Personal opinions will not allow us to understand with precision how MQA functions and the comparison results Dale brings up in his text are his opinions based on personal experiences and cannot be thought of as universal. For example, when he (and subjective reviewers) says something sounds "darker", this is a subjective quality which doesn't have a precise definition, likely would need trained listeners to achieve agreement, and has no clear point of reference. This property of sound (qualia) could be experienced very differently for me if I were to hear that same passage. While it could be "true" for him, this is not a "fact" that applies to everyone nor is it necessarily a property of MQA itself (ie. we need to be sure that what was heard was an apples-to-apples comparison, the only difference being MQA processing).

We could still try to be objective with "measuring" human preferences. Again this involves control and rigor. For example, we can do blind testing with music known to be of high quality, the same mastering, A/B between the original 24-bit master vs. MQA decoding within the limits of making auditory comparisons, etc... This is of course the intent of my "Internet blind test" many months back.

I look forward to the results from McGill University's testing of MQA, curious to know what controlled methodology they employed and with what audio stimuli (eg. are we talking about test signals, actual music, what equipment, etc...)

Ultimately, it depends on what one is aiming for. For any of us, we can absolutely decide that all that matter is "what sounds good to me" (I think this is boulderskies' perspective). If so, then by all means the only testing needed is personal experience and one's own subjective opinion rules supreme. In this context, it really doesn't matter what levels of distortions are present or if concepts like "fidelity" or "accuracy" to the original encoded signal were achieved in playback. It's important however to have insight and acknowledge that the generalizability of such opinions could likely be limited to oneself...

Unlike subjective preferences, objective facts can be replicated by others and often can be accomplished with different methods. Facts are "universal" (within reasonable context of course).

There is finally one last important difference I'll point out between the subjective and objective. As powerful as objective facts can be, they can also be proven wrong. That is the intrinsic "risk" of claiming generalizable truth. We cannot prove opinions wrong because they are by nature idiosyncratic - "true" to the individual. Whether a man or woman prefers "the sound of MQA" is along the same lines as "what is your favourite music genre?". No point arguing about this because in a free society, we can accept a multitude of opinions.

As humans, for any concept, both objective facts and subjective opinions (eg. "feelings about that fact") need to coexist at some level when it's something that matters to us. Who knows, in the days ahead, maybe MQA can show us in controlled tests that "most people" prefer the sound of processing through the MQA encoder/decoder; some evidence of subjective preference in its favour. But this will not diminish the existence of measurable distortions; it just means that the levels of the distortions are below audible thresholds of objection or some people might like them (we can say the same for preferences around vinyl playback or tube devices that add distortion).

None of this changes the non-audio factual findings like the presence of the cryptographic signature embedded in MQA files. This is of course a separate entity using the knowledge of which we can form other subjective opinions...

-------------------------------

Have a great week ahead everyone :-).

Enjoy the music...

BTW: If you're looking for another high image quality UHD Blu-Ray movie to check out - Murder On The Orient Express looks great! Originally filmed in 70mm (65/5-perf).