Looking for general advice re interconnects

[BenLaw]

I'm probably being very simplistic, but I wonder how the example from horizon of hearing 'fa' or 'ba' depending on how you see the lips moving fits into the notion that you can train yourself to take into account (and ameliorate) expectation bias. They say on that programme that even though you know the sound is the same, you hear two different sounds depending on the accompanying visuals. My experience accords with that, so I am seemingly unable to change my perception to take into account expectation bias.

 

[matt49]

I'm not interested in an academic discussion ...

In which case we're arguing from different premises.

There are a fair number of blind tests that have been carried out under 'robust experimental conditions' and subjected to rigorous statistical analysis, and serious debate, see here for example: http://www.head-fi.org/t/486598/testing-audiophile-claims-and-myths

We're talking about completely different standards of robustness (in the same way as I'm interested in an academic debate and you're not).

But even this is far beyond the reach or the need of ordinary enthusiasts, almost any kind of blind testing, so long as care is taken to match listening levels, will teach the listeners a lot about what they are hearing. It can be quite chastening to to discover that you cannot tell one cable (or amplifier, or DAC) from another despite 'knowing' that the differences are 'night and day'.

+ 1

Some people resolutely refuse to take part, knowing they will be shown up, others refuse out of shear ignorance ('I know what I am hearing') but to me this is the crux of the thread, reality and opinion do not make easy bedfellows.

+ 1

Another possible factor is the economics of the hi-fi business. Wine journalists and critics do blind tastings all the time; it's simply part of the job. These days most of them have formal qualifications, e.g. the Master of Wine exams contain very rigorous tests in blind tasting. I've seen these people in action may times: they're really good. The fine wine business can support this extensive training because it has the money. High-end hi-fi is tiny by comparison. I'd hazard a guess that the toal sales of high-end hi-fi in any year are roughly the same as the value of wine produced by 2 or 3 top Bordeaux chateaux.

 

[relocated]

Good choice of cables (I use generic pro XLRs). One word of caution, if the system sounds different at home from the demo room it will almost ccertainly not be because of the cables. The different room acoustics, positioning etc will have a far greater influence. So don't be tempted to try and 'correct' any deficiency by spending big on cables. Instead, move the speakers, turn the speakers, move your seat and consider acoustic treatment.

Back to the OP. The above advice is absolutely essential. It is far more important than cables.

The 10% thing has been around for decades from when hifi mags percentaged out your spend re source, amp, speakers. Nowadays with didital it is pretty easy. You are almost certainly never going to tell the difference between digital cables so if you spend very little money it won't matter. So long as it fits properly, isn't too long or flimsy then it will be fine.

 

[davedotco]

Sorry, did not mean to sound so blunt, I am interested in any discussion that does not turn into a p*ssing competetion and fully accept that your expertise in psycology is sufficient to render any discussion we try have on the subject pointless. I have learnt a couple of things (from your references) and that is always a plus.

I would take issue with your comments about the robustness of blind testing, many are not taken that seriously, fair enough, but some are and the results subject to serious statistical analysis. A match for many of the wine taste tests you mention.

There are of course many instances of supposed wine experts being fooled by blind testing, non experts are fooled all the time and the reports of such testing are easily found.

Perhaps not the best example, though your suggestion that people can learn about tasting wine (during a course of study, say) such that they can pass blind tests is interesting. Perhaps we should have such a course for hifi enthusiasts.....

 

[matt49]

Sorry, did not mean to sound so blunt, I am interested in any discussion that does not turn into a p*ssing competetion and fully accept that your expertise in psycology is sufficient to render any discussion we try have on the subject pointless. I have learnt a couple of things (from your references) and that is always a plus.

The apology is thoughtful but unnecessary. No need to apologise for bluntness (not to me anyway). This is a forum after all ...

I likewise have benefited from technical (acoustic) engineering expertise that you and others on this site have. One good reason to come back here.

From a link in one of your previous posts I got to this mildly interesting paper, which is quite relevant to the OP's question about interconnects:

http://www.apiguide.net/04actu/04mus...teractions.pdf

The conclusions suggest that different cable designs do yield different electronic measurements. Whether that converts to differences in SQ is another matter. But it certainly doesn't give much support to the idea that we should spend much money on cables.

I would take issue with your comments about the robustness of blind testing, many are not taken that seriously, fair enough, but some are and the results subject to serious statistical analysis. A match for many of the wine taste tests you mention.

Yes, I didn't mean to suggest wine blind tastings were inherently more robust than hi-fi blind tests. What I meant was that there exists a large number of wine experts who do perform very well in blind tastings, and that one reason for this might be the amount of time they spend doing it.

There are of course many instances of supposed wine experts being fooled by blind testing, non experts are fooled all the time and the reports of such testing are easily found.

True.

Perhaps not the best example, though your suggestion that people can learn about tasting wine (during a course of study, say) such that they can pass blind tests is interesting. Perhaps we should have such a course for hifi enthusiasts.....

That would be a real step forward, though I'm not optimistic about it ever happening. Who would fund it?

 

[matt49]

The suggestion that cables should make up 10% of the system cost clearly came from a cable company.....

I think it is upto 10%, not 10%.

There does seem to be a common figure. Check out 'Chapter 4' http://www.whathifi.com/video/get-the-best-from-your-hi-fi

The 10% thing seems to have become semi-institutionalized. When I bought an amp and DAC several months ago, the retailer was doing an offer: free interconnects up to the value of 10% of your spend on other kit. Needless to say, I took them up on the offer, though it might have been fun to say "You can keep your fancy interconnects, they don't make any difference!", just to see the look on their faces. (Of course, I did try the 'I'd rather have some money off the amp and DAC" line, but Head Office wouldn't have it.)

 

[BigH]

The suggestion that cables should make up 10% of the system cost clearly came from a cable company.....

I think it is upto 10%, not 10%.

There does seem to be a common figure. Check out 'Chapter 4' http://www.whathifi.com/video/get-the-best-from-your-hi-fi

The 10% thing seems to have become semi-institutionalized. When I bought an amp and DAC several months ago, the retailer was doing an offer: free interconnects up to the value of 10% of your spend on other kit. Needless to say, I took them up on the offer, though it might have been fun to say "You can keep your fancy interconnects, they don't make any difference!", just to see the look on their faces. (Of course, I did try the 'I'd rather have some money off the amp and DAC" line, but Head Office wouldn't have it.)

Probably because the mark up on cables is much higher than the actual gear.

 

[matt49]

The suggestion that cables should make up 10% of the system cost clearly came from a cable company.....

I think it is upto 10%, not 10%.

There does seem to be a common figure. Check out 'Chapter 4' http://www.whathifi.com/video/get-the-best-from-your-hi-fi

The 10% thing seems to have become semi-institutionalized. When I bought an amp and DAC several months ago, the retailer was doing an offer: free interconnects up to the value of 10% of your spend on other kit. Needless to say, I took them up on the offer, though it might have been fun to say "You can keep your fancy interconnects, they don't make any difference!", just to see the look on their faces. (Of course, I did try the 'I'd rather have some money off the amp and DAC" line, but Head Office wouldn't have it.)

Probably because the mark up on cables is much higher than the actual gear.

In other words, not only is the SQ benefit dubious, the stuff is also being overpriced.

 

[AEJim]

On the subject of testing components in a system - expectation bias is very difficult to overcome, blind testing is a solution but is awkward and requires assistance.

What I would recommend as a simple alternative is A-B-A testing, ie: Listen to your system, swap in new component and listen again but then go back to the original and listen again. It's quite an effective and quick way of gauging how real a difference may be. Often when making a change you hear a "definite improvement" initially but when you swap back it's less obvious in hindsight. It doesn't completely get around expectation bias but it is far better than a simple single swap for new component. Also use a track you know well and listen for a short period, you know quite quickly if something is right or not. When happy with your system you can always have a longer listen with more material.

The difficult thing with psychoacoustics and expectation bias is that if you think a component is better for whatever reason, even if technically it can make no difference, then in a very real way it IS better for YOU - you will hear the improvement! People will argue that there is technically no difference and may well be correct, but if you believe there to improvement then that improvement will be real to you. It's the reason there are so many ongoing debates on Hi-Fi forums about cables etc and it's an impossible battle to win either way because both camps are right in their own way.

 

[MajorFubar]

My view is I can accept the theory that different cables affect (analogue) sound, but what I can't see is the correlation between a cable's SQ and price. Or put it another way, what is it about the construction of expensive cables which we're lead to believe makes them sound better? Or put it even another way, what is it about good-sounding cables which we're lead to believe makes them expensive to manufacture?

I can look at an expensive hand-built turntable, I can see and feel the quality of the build and workmanship, I understand why that quality of build costs money, and then I understand the basics of why that high-quality solid build potentially has a positive affect on the tt's SQ. But not so a cable. I can't see what's in a brilliant-sounding cable which makes it more expensive to manufacture than a cheap one.

 

[davedotco]

The API guide linked in Matt49's thread gives an excellent explanation of how (speaker) cables can and do measure differently and more importantly how different amplifier and speaker characteristics affect the cables performance. Some of the HF roll off in some of the samples were pretty shocking.

Sadly it is beyond the brief of the paper to investigate how these changes relate to what is actually heard but I think the real lesson here is how the cable interacts with the components on either end.

Speaker cables (in this case) are clearly system dependent, something that was obvious enough in my time as a dealer, but rarely analysed.

AEJim's suggestion of A - B - A testing is sensible enough and many of us (more anal types) have been doing it for years but even so it really no substitute for blind testing.

I strongly urge any enthusiast to take part in a blind test if given the chance. If nothing else it will give you a really good handle on how small differences in suposedly very different components can be.

 

[BigH]

Yes I agree with AEJim and have suggested something similar, I have read several reports along the lines of changing the cable to a more expensive and it sounded better, then doing the same again with a more expensive one and so on but then returning to the original and he could not hear any difference. Maybe on the first the amp was just warming up or his brain was. I agree blind testing is better if you can arrange. Sometimes people can hear a difference but they don't believe the difference is worth the extra cost, sometimes the more expensive cables sound worse.

 

[matt49]

Sadly it is beyond the brief of the paper to investigate how these changes relate to what is actually heard

For that you'd need a psychologist ...

 

[davedotco]

Sadly it is beyond the brief of the paper to investigate how these changes relate to what is actually heard

For that you'd need a psychologist ...

You will need a statician too, the bastards get in everywhere........ :hand:

 

[matt49]

Sadly it is beyond the brief of the paper to investigate how these changes relate to what is actually heard

For that you'd need a psychologist ...

You will need a statician too, the bastards get in everywhere........ :hand:

But the psychologist will have been trained in statistics, so you won't need the EDITED statistician. 8)

 

[busb]

I just going to keep asking until someone give a satisfactory answer - What proof is there that double blind ABX testing works for audio?

 

[matt49]

I just going to keep asking until someone give a satisfactory answer - What proof is there that double blind ABX testing works for audio?

It depends what you mean by "works".

The ABX testing that's been done on hifi kit has in most cases produced statistically null results, i.e. the subjects in the tests have displayed preferences (e.g for an expensive amp as against a cheap amp) that are statistically no different from complete randomization.

But (and davedotco may disagree here) this ABX testing has not been carried out to standards that those in the field of psychoacoustics would regard as academically robust.

Also there are vanishingly few hifi "experts" who are willing to submit themselves to these trials, for a variety of reasons.

The conclusion I draw from this is that the jury is still out and is likely to be out for a long time yet.

That may or may not be a satisfactory answer. :quest:

 

[davedotco]

Double blind and ABX testing are two different techniques, i believe they are quite different in psycological terms, thogh matt49 will put us straight on that.

In both cases the test equipment is set up in such a way that all variables are excluded, apart from the the items being compared, ie in the case of an amplifier test the tone controls would be set flat, the output levels carefully matched and care taken that both amplifiers are working well within their designed capabilities.

In a double blind test the test subjects are played excerts of music on one system, then on the other. They are simply required to say which one they think sounds best. Importantly, the person carying out the test does not know which amplifier is which at any point.

An ABX test is subtly different, in this case music is played on one system then the other, then a third time on either the first or the second system. The subject as asked whether the third sample is a repeat of the first or the second sample.

This is repeated as often as possible, with as many subjects as possible and the results calculated statistically, but I believe differently in the two cases.

There are also variations on these tests where tests are inserted where the two excerts are played on the same system. ie there should be no difference, and the results of these tests analysed too. All of these results can be (and have been) subjected to rigorous statistical analysis.

All of this analysis is of interest to academics but the ordinary hifi enthusiast simply wants to know whether he can tell a difference or not and the answer is often not.

Perhaps more importantly the enthusiast gets to see just how tiny the differences can be between, apparently, quite different products.

 

[matt49]

Double blind and ABX testing are two different techniques, i believe they are quite different in psycological terms, thogh matt49 will put us straight on that.

That's an invitation if ever I saw one, but I fear I would bore everyone (if I haven't already). Well, OK then, just a quickie.

All serious testing is double blind. Double blind simply means that both the subjects of the test and the people conducting it are ignorant as to the identity of the things being tested. This is so that the people conducting the test can't convey any subliminal cues to the subjects.

ABX testing is one of a large number of (double blind) sensory evaluation test protocols. Davedotco has described it well above. There's a brief and accessible description of various test protocols in Meilgaard, Carr, and Civille: Sensory Evaluation Techniques, pp. 59-98. Sensory evaluation tests are big business in e.g. the food and cosmetics industries. Their methodologies and results are controversial. AFAIK no serious studies have been done on sensory evaluation techniques for hi-fi.

 

[busb]

I just going to keep asking until someone give a satisfactory answer - What proof is there that double blind ABX testing works for audio?

It depends what you mean by "works".

The ABX testing that's been done on hifi kit has in most cases produced statistically null results, i.e. the subjects in the tests have displayed preferences (e.g for an expensive amp as against a cheap amp) that are statistically no different from complete randomization.

But (and davedotco may disagree here) this ABX testing has not been carried out to standards that those in the field of psychoacoustics would regard as academically robust.

Also there are vanishingly few hifi "experts" who are willing to submit themselves to these trials, for a variety of reasons.

The conclusion I draw from this is that the jury is still out and is likely to be out for a long time yet.

That may or may not be a satisfactory answer. :quest:

What I find so curious is that they get mentioned as being absolutely foolproof. I see two problems: firstly, reliance on short-term memory. We can assume instant switching so no delay but what evidence is there that that we can actually remember enough to draw any conclusions? Audio introduces some particular problems. For instance if we want to compare the quality of two different photographic lenses we could print out the results from each & examine both side by side, looking for particular aspects such as sharpness or colour fringing at both the centre & around the edges - we can scan with our eyes, notice aspects, recheck etc. We are not constrained to only look at one or the other in sequential time frames.

With audio, we have to listen, then memorise then listen again - so what are we testing - audio memory or potential differences? Any differences noted are totally reliant on memory - if testing amplifiers, no one is suggesting we level match with one playing through the left an the other through the right speaker simultaneously! Am I alone in wondering whether or not short-term memory is better than longer-term memory? I ask because I find I get very confused with short quick-fire tests! I also often notice differences in perceived SQ after months not seconds. This may occur after playing a particular piece where I notice additional details. If I’m not listening out for differences & am just enjoying music, Expectation Bias is surely not an issue. Short-term memory maybe very convenient for formal DB ABX testers but only if effective!

So just how we confirm short-term audio memory is useful or not? ABX is very effective at filtering out false positives by the very nature of its statistical analysis but surely any methodology should be good at weeding out false negatives?

Let’s assume for illustration that we have room full of sceptics who firmly believe that nearly all differences with amplifiers are imagined. So they either genuinely cannot tell or have an axe to grind or just plain old Expectation bias plays its part! Once this particular group's responses have been analysed the numbers are close to 50 when averaged. This result would be similar to a random result. If a similar test was carried out on (for the sake of argument) a group who could tell the differences between amplifiers under test, the average result would be higher than random where the greater the number, the more statistically significant would be the results. As far as I know, most such tests "prove" no differences can be heard & the statistics are the same as being random. OK so far - still with me? Do you see a flaw to this like I do?

How do we differentiate between there being no detectable differences because none exist or the possible fact that the test method is flawed & that no subjects can tell anything whether or not they are real. What I'm saying is that a group of near deaf people could take the test & the result would be random unless we can filter out the potential false negatives. Their results would hover around 50 but never be lower. Unless I'm failing to understand the methodology of such testing we need to have some sort of control during the test such as real repeatable & deliberately introduced differences. These could be left right level imbalances, added distortion etc. The purpose of these controls is to highlight the deaf or axe grinders so no one can skew the results therefore making the results either less valid of highlighting the flaws of the methoditself. The level of these deliberately introduced controls could be incrementally increased to the point where they should be obvious. If they can't, be heard, what hope is there that the rather subtle variations in equipment can be heard? Without these controls, I fail to understand how anyone can think the tests work. If we sort out the potential false negative problem, we also get a handle on whether or not ABX tests prove anything. I'm sorry this is very long-winded but I can think of a simpler way of explaining it!

 

[BenLaw]

Last time you asked the same question I linked to a number of ABX tests which had been passed (eg for speakers). So that rather disproves your notion that a segment of people deliberately always fail them.

 

[davedotco]

Once again I find the discussion getting bogged down in detail and circular arguments.

By bringing up and supporting blind testing I was attempting to share some experiences where, clear and obvious differences, clearly observable in sighted tests largely disappear when the listening is blind.

Any reasonably competent blind test will show this and in this context the rigorous analysis of results does not matter, it is simply that most people are shocked by how small these differences become when visual (and perhaps other clues) are removed, I must admit to being thoroughly chastened after my first experience of such testing.

While the science undoubtably matters to academics, for me it is the experience of the listeners that really is to the point. Irrespective of the results the difficulty in hearing differences blind, that are considered 'night and day' in sighted tests should, at the very least, teach that subjective, sighted testing is not to be relied on.

Matt49 - If I impliesd that ABX testing is not double blind, then I was being unclear. As you say double blind is essential in any serious test.

 

[relocated]

I just going to keep asking until someone give a satisfactory answer - What proof is there that double blind ABX testing works for audio?

It depends what you mean by "works".

The ABX testing that's been done on hifi kit has in most cases produced statistically null results, i.e. the subjects in the tests have displayed preferences (e.g for an expensive amp as against a cheap amp) that are statistically no different from complete randomization.

But (and davedotco may disagree here) this ABX testing has not been carried out to standards that those in the field of psychoacoustics would regard as academically robust.

Also there are vanishingly few hifi "experts" who are willing to submit themselves to these trials, for a variety of reasons.

The conclusion I draw from this is that the jury is still out and is likely to be out for a long time yet.

That may or may not be a satisfactory answer. :quest:

I would respectfully suggest that the overwhelming reason is that, 'the golden eared' brigade are just too scared to find out that they are not so golden eared after all. Career over for reviewers, shown to be unable to tell the difference between different equipment and cables and not a whole lot of mileage in participating in the purchase [or otherwise] of publications that use such people.

No point in testing people who do not believe that differences exist because, they will deliberately not hear any difference or are actually hearing challenged and don't know it. A couple of reviewers have 'outed' themselves recently about subjective reviewing, earning a living and giving their readership what they want. The vitriol expressed by people was a disgrace. The truth, even though mildly veiled, was too much for the audiophiles to cope with.

 

[CnoEvil]

I have linked to this before, which is a blind test of cables....and basically shows that they were able to tell the best and worst from each other: http://www.nordost.com/default/pdf/hifiplus_issue34.pdf

If the link stops working, go to Nordost.com -> Reviews -> Speaker Cables listening tests

 

[busb]

Last time you asked the same question I linked to a number of ABX tests which had been passed (eg for speakers). So that rather disproves your notion that a segment of people deliberately always fail them.

Ben

If you read this, I apologise. If you link to the tests, I'll read them, I promise.