Cable Pathways Between Audio Components Can Affect Perceived Sound Quality

[fas42]

Pleasing to see this author continuing his efforts to unearth "what's really going on!" in papers - a properly done bit of research, as compared to most efforts aimed at debunking audiophile concerns; with plenty of quotable quotes within the body of the writing.

[fas42]

Well, the conclusion very clearly states what the intention was,

 

Quote

High-end audio is a subject that is shrouded in controversy.
Aside from loudspeakers, consumers exhibit varying
degrees of skepticism as to what affects sonic performance.
The most contentious ingredient in the chain is the
interconnection between components, which concerns both
the topology (balanced versus single-ended) and the
characteristics of the cable itself. This work shows that two
system configurations differing only by the interconnect
pathway are audibly discernable, even by average listeners
with no special experience in music or audio.

 

Right, he did an experiment to test that, and got significant results ... next, please ...

[fas42]

... And all the excuses start spilling out ... when there's an agenda, never let a single thing past, just in case it could do some damage to one's magnificent edifice of thinking, 😉.

[fas42]

25 minutes ago, The Computer Audiophile said:

I’m looking for more info to support or disprove what I thought the paper was designed to prove. 
 

Can you help?

 

The author says, in the introduction,

 

Quote

Nevertheless, to the author’s knowledge, proofs of
audibility through direct listening comparisons where the
cable pathway was actually changed (i.e., not simulations
or theoretical estimations) have not been previously
published.

 

Therefore, doing the carefully controlled test to verify this, and publish the results, is fully justified.

 

As I implied earlier, we now have adequate research, that is the first step. Next step, don't change the pathway type, and only change the cable itself ... and repeat the testing - we're moving towards developing a testing regime that is good enough to elicit an acceptable result when testing, "subjective things".

[fas42]

32 minutes ago, Iving said:

These are my notes reading the paper once thru this morning. imo the paper’s chief shortcoming is that unambiguous Design is not evident from soup (Abstract) to nuts (Conclusions) as it ought to be.

 

Ah, we need this to be the "perfect" scientific paper, to even begin to break down the impasse between the subjectivists and objectivists, in some tiny way ... my apologies for not appreciating this critical need, 🙂. Of course, it's far more exciting and entertaining to witness ongoing decades of pointless biffo between the two camps, getting everyone nowhere - so, who was the fool who wanted a solution to this silliness ... ?

 

32 minutes ago, Iving said:

In the Abstract, the author immediately introduces a study confound; i.e., “cables and topology”. “Cables” is fairly specific. “Topology” is not. I am imagining that “Cables” is an easily constituted IV with levels, where each level corresponds to a particular actual cable or cable type with attributes to be specified. “Topology” is suggestive of either further IVs (rendering the Design more sophisticated) or vagaries = nuisance variables (which would demand statistical control of co-variates). In any experiment you just have to isolate the effect you wish to investigate. Where more than one Independent Variable is implicated it is *necessary first* to report/interpret the interaction between them since, if it is statistically sig., you can’t interpret Main Effects without qualification. The question in the reader’s mind is whether “cables and topology” involves multiple IVs, or whether the experiment should be more sensibly carried out in phases. By the end of the Abstract, the Author has not de-confounded “two different analog-interconnect pathways” wrt “cables and topology". We shall have to assume we have one IV with two levels (the cable routes).

 

I like that the Introduction takes on SSC vs. EMP. Somewhat aghast at the public spectacle that was the recently the now-closed ASR “We have a volunteer” thread, I watched Amir’s SSC-hinged video on Blind Tests and Trained Listening. I know there’s something I want to say about it – the SSC issue, but haven’t been motivated to work that out on paper (as argument) to my satisfaction. I am going to gloss over the Blind Test principles the Author here mentions, since the difficulties putting together an unassailable (i.e. ruling out rogue explanations for a pass) audio Blind Test are monumental as demonstrated in the ASR "We have a volunteer" thread just mentioned. My consistent stated case there was that the project was anyway doomed from the outset because of “combative nonsense” between Amir and @GoldenOne. I shouldn’t be surprised if Amir were old enough to be Golden’s Grandpaw! - and he ought to know better. By the end of the present “cables and topology” paper’s Introduction, the confounding issue is exacerbated since we now have mention of signal simulations and grounding. I say “exacerbated” because, rather than opening new horizons, the reader wants to be taken down a funnel from the unspecific (misc. related ideas) to the specific (clear Design stipulating vars., levels of those vars., DV, anticipated relationship inc. statistical tests germane).

 

The equipment section depicts the system as Server > DAC > Power Amp > Speakers. We are concerned with interconnects but which. There's an AES/EBU digital interconnect Server > DAC. OK. Doesn’t say whether this is variable. Yes – Bal *and* Unbal interconnects of different makes/lengths from DAC to Amp with implication that this difference constitutes 2 levels of an IV. The problem is that without refinement *everything* that is different between the Bal/Unbal routes can explain changes in the Dependent Variable (which we don’t know yet but presumably is a SQ index of some kind). OK – speakers bi-wired. No mention of speaker wiring as variable. After a little audiophile masturbation (in which the attributes of the testing apparatus and room are extolled), the stated aim becomes clear: “In the hope that some audible difference will be detected, the cables in the two configurations were chosen to be as different as possible. With its much shorter length, balanced (versus single-ended) topology, and faster dielectric, the higher-end cable A can be expected to have a more detailed and accurate sound. The question was whether interchanging these interconnects would produce a recognizable and memorable timbral change that would be discernible in blind listening tests.” In other words, the Author is *optimising* the prospects of revealing *any* SQ delta between the two cable routes. The Author is *not* trying to establish that it is a *single* aspect of the route that can be discerned SQ-wise (Bal vs. UnBal; Short vs. Long cable; particular manufacturer; simulated vs. real music; grounded vs. not grounded etc). I find this undisciplined as a scientific study because of messy Design. It is more an attempt to fire a shot across the bow of “You can’t hear cables.”

 

Section 1.2 introduces a mic/amp/oscilloscope as listener. I wasn’t expecting this. Usually an Abstract – certainly if not the Introduction - would have avoided this kind of surprise for the reader. Most of the section speaks to sensitivity. As far as I could see on one read, this section was about explaining that the room was good enough to enable human listeners to make sensitive judgements.

 

Section 2 revisits EMP vs SSC. So we are back to human listening . The lh column p.5, to my eye anyway, betrays that the Author is not a psychologist – but I don’t want to quibble. Section 2.2 justifies the Blind Procedure. At first I don’t worry too much since any statistically sig. effects could be understood in the context of these details as well as cable confounds (all getting a bit motley even so). Now we are reading about non-audiophile listeners being familiarised with audiophile vocabulary which presumably is meant as a prime - not sure about that and its implications (for any DV) without further contemplation. OK so this pilot involved 3 listeners, and then we have 18 human listeners for the test proper. The DV is Cable (route) A/B or B/A *order judged correctly*.  The 18 human listeners comprised 11 female / 7 males – fair enough if sex is not an IV. But  then we have a mix of musicians [7] and non-musicians [11] which has to be a covariate (with 2 levels). But we have not seen any formal Design yet.

 

Notwithstanding we are presented with “Results” in 2.3. Now we are told that subjects come and go/provide varying degrees of co-operation inc. wrt # data-yielding trials. The object seems to have been to procure trials. Of 59 trials, 43 were judged correctly (AB or BA order cable route) and 16 not judged correctly. To the extent that we can accept such results at face value the statistical sig. needs barely confirmation. There is a suggestion of a one-tailed test which is not great even if only because two-tailed tests are almost invariably used in conservative studies until causal relationships can be ruled out – but doesn’t really matter. The Author has used a non-parametric Wilcoxon test. Cable routes 1 & 2 are the levels of the IV. Wilcoxon is “within-subjects” meaning participants generated data for both levels of the IV. There is far too much info presented here for a simple test with so few Ss, and the sig. value for p is being oversold. It’s like reading Amir boasting about his blind test performance! The Author is saying that sub-tests suggest no trend in performance across trials i.e. what is tantamount to practice or training effects. I’m not going to comment on these sub-stats because I don’t think the Design foundations are strong enough – let alone df wiggle room and so on. Now we have a Ch-Sq. test too collapsing the Ss into categories for succeed/did not succeed. Undisciplined underlining imo. A formal Design would have determined a single, suitable statistical approach.

 

I’m not familiar with “phonemic restoration”. I would expect musicians to do better on any musical discernment task simply because of familiarity. The suggestion that they do worse on a task discriminating two cables where the second is inferior because their musicianship *obliges* them to “mask” deficits sounds like bunkum to me. Even if the effect were confirmed in a larger study where confounds were removed, I can think of other explanations. Anyway, the Author is not pretending the limited data here are authoritative.

 

Section 3 begins with the loose but broadly fair statement “If two audio configurations are audibly distinguishable, then physical differences between their signals must necessarily exist.” There follows great detail about the electrical characteristics of the respective cables. I was not at all surprised to read “interconnect A has dramatically lower noise than B”.

 

I do think that this study has been fastidious (egging the room and stats puddings) where it needn’t have been, and lacking in fastidiousness elsewhere (Design). @The Computer Audiophile is on beam with the apples & apples thing; i.e., there is too much confounding of differences in the two cable routes – many of them known and explicit in the study as reported.

 

The banal outcome, as I read it, is that there was a substantial and demonstrable *noise* diff. between two cables, and 18 listeners could hear it at a statistically sig. level. @idiot_savant got it in one even after a skim read. [Where did I miss “the screen was not correctly connected” - ?]

 

Notwithstanding, the Author makes the bold claim, “To the author’s knowledge, this may represent the smallest change in an audio system proven to be discernible through IRB approved blind listening tests.”

 

If there's anything worth discussing further here, it would be exactly how the noise differential in the two cable routes was caused/measured/heard. And then whether we are looking at anything ground-breaking.

 

Friendly word to Frank @fas42. All research papers are not so different from legal opinions. Some are good and some are weaker. There is no alternative to the fine-tooth comb. 😊

 

I think it's also about time to rip apart the Bible, you know - its a pretty ridiculous hodge podge of odds and ends of thoughts and experiences ...  can't be anything in there of any value to anyone, with all those flaws, 😉.

 

What's remarkable is the determination of just about everyone to make absolutely no movement forward in trying to get a better handle of what matters, and what is not so important, in making the playback of recordings a better experience. Well, for those whose need for such is so great, may you be blessed with that very outcome ...

[fas42]

7 hours ago, pkane2001 said:

Fully balanced and single-ended signal paths are sufficiently different internally that any differences could be due to electronics rather than interconnects. The author is, unwittingly, changing too many variables by comparing balanced to single-ended interconnects to make any conclusions about just the interconnects. Different components, different signal path, different processing of the signal between these two cases, possibly different signal levels at the output. All of these are important to eliminate as possible confounds before declaring that interconnects cause the differences. 

 

Also, the fact that balanced vs. single-ended can produce different levels of noise at the output is not surprising, since these were created specifically to deal with noise.

 

I do wonder whether anyone has actually read this paper - just try perusing the Conclusions section, okay?

 

When people have an agenda, all forces that can be applied will be applied, no matter how feeble they are ...

[fas42]

Strangely, the concept is that unless "perfect" integrity is the nature of the chain, that there will be audible differences - in the real world everyone knows this, and it forms the backbone of what everyone does - yet, a formal paper investigates, and puts this in writing, and people fall to the floor in amazement 😉 ... the human condition is quite remarkable to observe, at times ... 🙂.

[fas42]

15 minutes ago, pkane2001 said:

 

Understanding why and how things work is my only agenda, Frank. I understand that you just naturally come by the understanding and knowledge of how things and work and how to improve them. Not all of us are this gifted, so some of us actually need to ask relevant questions. Okay?

 

 

No. First step, acknowledge that there is "something going on" - don't apply maximum effort to convince yourself that that you were deluding yourself, by doing measurements until absolutely "nothing to be seen here" - in some parts, they call this being open minded; and they don't let their brains fall out, BTW, when doing so, 😉.

 

The "why" and "how" comes later - accept the WTF'ness of the situation, and proceed with the realisation that something is happening, and should be dealt with ...

[fas42]

Just now, pkane2001 said:

 

Not how science works, I'm afraid. We are discussing a scientific paper that purports to describe a scientific experiment.

 

Love the way you throw in the word, "purports" - ahh, that science find it so easy to rise above the emotions that everyday folk succumb to ... 😜