USB Cable Poll: Redux

[Alan Sircom]

You said company which I took to be your employer hence the business about publicity for your magazine when you passed. The fact you meant cable companies and the importance you seem to place on this relationship is rather telling. Whatever, it is good to have input from people such as yourself on forums like this. I am off for my dinner.

If a company's product is being put up for test that might cost them business whatever the outcome, it would be rude not to discuss it with them. That and the whole potential loss of business law suit that might ensue as a result of me trying to make an extra million for my millionaire boss.

My company would probably say 'yes', then 'no', then 'yes' again, then 'definitely not'. Given my oppo in The Absolute Sound thinks blind tests are so wrong as to be almost immoral, putting us up for a test that relies on blind testing wouldn't be taken well, whatever the outcome.

 

[Alan Sircom]

So you tested cables that you cannot name along with fellows that you also cannot name, and you have no documentation or interested spectators to reassure us that the results were in any way meaningful.

With all due respect, why on earth would you think people were stupid enough to accept that?

As for only taking a challenge (that could make you $1M) only with the agreement of the companies that refuse to provide you with the means to repair your shed, again, there's a credibility issue there for me I'm afraid.

They were not published for your edification. They were an in-house evaluation that I covered in a column a few years ago. I mentioned them in passing. If you want the details, it was a generic cable, two prototype cables from AudioQuest, a prototype cable from Crystal Cable, a prototype cable from Nordost and the then-new Cardas Clear USB. Most of the prototypes differed from the end user cables, in some cases fairly substantially. Test subjects were Chris Binns, Chris Thomas, Roy Gregory and several members of a US forum who did not wish their identities disclosed. As it was to evaluate whether there were differences in cables rather than necessarily picking out winners, I respected their anonymity, until now.

Deal with this in whatever way you want to.

JREF demands full compliance from all parties. I can't randomly decide to take someone's cables and put them to the test, even if I were to ride roughshod over the company involved. Randii would just get in touch with the company and the test would be over. And I would get the blame. That's what happened with Michael Fremer. So, I'd have to do it in the full knowledge and acceptance of the cable company. And none of them want anything to do with it, because of their distributors in the far east, who think Randii is a trickster.

It wouldn't make me $1m anyway. I work under quasi-Texan employment rules. What's theirs is theirs. What's mine is theirs too.

 

[Alan Sircom]

I'd be stunned, but, if he got it right say, ten times in a row, then would that not strongly suggest there really were audible differences?



I don't know how many times something would need to be correctly differentiated to actually prove something, but I don't see what issues there may be? If one can identify something as mentioned above, even ten times in a row, then while nothing is definitively proven, it strongly suggests audible differences.

If not and the test procedure is well controlled, then I think it strongly suggests no audible differences.

That's mostly irrelevant anyway as nobody ever has reliably differentiated between things that science dictates do not differ, under controlled conditions, so why not just trust the science in the first place?

To achieve 95% statistical confidence in a test, you would need to repeat the test ideally at least 10 times and the test subject would need to correctly identify the device under test at least eight times out of the possible 10. You would be able to conclude at this point that it was very, very likely the test subject wasn't just guessing. More test repeats are better, although past 25 tests per subject, it gets academic.

However, it would also be statistically significant if a test subject got a test right just once out of a possible ten tests. This would imply the test subject was getting it wrong, but getting it wrong with 95% statistical confidence. This is usually dismissed as an outlier.

You need more than one test subject, too. You need dozens to limit the possibility of false positives and negatives.

 

[Joe P]

Max,

That's mostly irrelevant anyway as nobody ever has reliably differentiated between things that science dictates do not differ, under controlled conditions, so why not just trust the science in the first place?

You're putting the science cart before the empirical horse. You start with a hypothesis you want to test, a hypothesis that can be derived from what you expect from theory or what has been observed, then you design an experiment to test it.

The data drive the science, not the other way around. You often come to these debates from the other side: science says they can't be different, so they're not different.

I agree that some things just don't seem worth the trouble to test, but if you're being scientific about this you have to be an good empiricist.

Joe

 

[maxflinn]

Max,


You're putting the science cart before the empirical horse. You start with a hypothesis you want to test, a hypothesis that can be derived from what you expect from theory or what has been observed, then you design an experiment to test it.

The data drive the science, not the other way around. You often come to these debates from the other side: science says they can't be different, so they're not different.

I agree that some things just don't seem worth the trouble to test, but if you're being scientific about this you have to be an good empiricist.

Joe

Joe, fair points, what I should have said was why keep on blind testing things that science dictates do not differ.

Having said that, it obviously isn't a coincidence that nobody has, to my knowledge, proved science wrong by way of blind tests involving audio products.

FWIW and IMO, most of the things Darren listed may well differ audibly to certain degrees, but digital cables? Impossible.

 

[maxflinn]

Anyway, Joe, what's far more important is that I've recently discovered Morcheeba, and they're absolutely brilliant.

http://www.youtube.com/watch?v=VyRxTTegjXM

 

[andrewd]

And how will that affect the audio on a DAC with a ASR or other means of generating the audio clock independently of the incoming clock?

Just as the precise voltage level of an incoming bit doesn't matter as long as it stays in the correct tolerance band defined for a "0" or "1", the outgoing timing can be completely regenerated/resynchronized so that the precise timing of incoming data doesn't affect the data being read as long as it stays within the bit window.

The designers of ASR admit that it is not a panacea, and this is backed up by the observation that even DACs with ASR can benefit from lower jitter on the input bitstream.

 

[chiily]

Sure:

We submitted six USB cables to a blind test before the whole USB cable thing gained traction....

This was performed as a panel test, rather than a series of solo tests. One particularly strong listener managed to correctly identify brands (despite not knowing which brands were submitted) in three cases, due to them having what he considered to be a 'house' sound of the relevant brands.

'House' sound for a USB cable? I need to shake that cable's designer by the hand.

How on earth can a house sound be designed into a USB cable?

Someone will come along soon to sell me a humidifier to maintain the perfect humidity in my house to ensure correct propagation of the wifi signal...geeezzz...

...hold on, there must be a market there ... I wonder if different mains leads for the humidifiers will improve the sound quality of wifi streamed audio...there is a new vista of foo to market ...

In fact by simply wetting the floors will improve propagation ...lucky hifi owners in Somerset, there is a silver lining to all this flooding. we'll see those forum posts any moment now, "How living on a flood plain improved the sound of my hifi"
Let me go to bed...

 

[maxflinn]

'House' sound for a USB cable? I need to shake that cable's designer by the hand.

I'd say you'd be shaking the hand of some factory owner in Zhongshan, Guangdong Province, China.

How on earth can a house sound be designed into a USB cable

If the chap in Zhongshan doesn't have his robots do it during production then I guess it must be secretly incorporated into the external casing in Blighty prior to sale, perhaps a super small but powerful, active, intelligent nano graphic equalizer which filters the frequencies using a method first used on the planet Zong that Joe learned about from Spock?

Looking forward to the trekkie and pony pics that will hopefully shed some more light

 

[Joe P]

Max,

If you watch this video, you'll see Captain Kirk rewiring the ship with fancy cryo-treated mains cable after handing Finney his arse on a platter.

Joe

 

[ItemAudio]

One of the reasons it's unwise to claim 'impossibility' for an outcome here relates to the reality of combinations 'in the field'.

Windows OS is a great example. It's not only that it's badly written; it's tackling a genuinely impossible task: catering for a moving target of billions of combinations of hardware and drivers - it will inevitably crash/fail/suffer unresolvable conflicts. I often think about the Challenger shuttle, too: engineering and QC of the highest calibre; O-rings perfectly fit for task . . . until they got cold - which rapidly butterflied into an unpredicted (though not unpredictable) terminal chain reaction.

Every listening room consists of unpredictably related parts; each DAC is differently susceptible to external power and jitter issues; each system is differently vulnerable and resolving; each listener is differently tuned for acuity. And this is borne out by the universal lack of agreement about what works best. How could it be otherwise?

So when there are concrete material differences between cables - and there's no difference in this respect between those carry high and low-frequency / square or sine signals - it's quite possible these will be audible.

Even this thread shows that differences between USB cables are audible to some - 50%, in fact. The more controversial issue is defining 'audible', which is properly the subject of another thread. This is only a listener survey, not an inquest.

 

[merlin]

Even this thread shows that differences between USB cables are audible to some - 50%, in fact. The more controversial issue is defining 'audible', which is properly the subject of another thread. This is only a listener survey, not an inquest.

More misrepresentation.

Very sad to see this day after day.

Look again at the figures. Only 22% claim to have heard differences - with no indication whatsoever as to whether these tests were subject to expectation bias - scientifically the overwhelmingly most probable reason for these people having this view.

 

[Julf]

The designers of ASR admit that it is not a panacea, and this is backed up by the observation that even DACs with ASR can benefit from lower jitter on the input bitstream.

Any pointers to controlled, verified tests showing that?

 

[Julf]

We submitted six USB cables to a blind test before the whole USB cable thing gained traction, to see if this was a category that justified investigation.

Tests were performed A-B, but slowly because of the difficulty of re-establishing connection with the DAC. Cables were presented in a random selection and repeats were run through the day. One of the cables was a freebie to act as benchmark.

This was performed as a panel test, rather than a series of solo tests. One particularly strong listener managed to correctly identify brands (despite not knowing which brands were submitted) in three cases, due to them having what he considered to be a 'house' sound of the relevant brands.

Alan,

Would it be possible to describe how the blinding was done - was it a single-blind or double-blind situation? If it was a panel, was there discussion (so that the opinion of one panelist could affect the opinions of others)? Would it be possible to publish the actual test notes (with names of brands and listeners anonymized)?

 

[ItemAudio]

I don't want to give the impression too strongly that blind ABX perception tests should be dismissed out of hand, but their reputation is badly distorted in the reflection of medical trials. Failure to recognise crucial dissimilarities leads to a blind alley.

Blind tests are conducted with various objectives: in this instance, the goal is to determine whether listeners project impressions of characteristic differences in the absence of genuinely differentiated external stimuli: are they 'hearing' things that aren't there?

However, when auditioning USB cables, A≠B. The listener is the object of the test. It's a 'sword in the stone' scenario: a single act of discrimination proves once and forever that two cables produce genuinely differing output. Subsequent blind trials are irrelevant: they're no longer about whether the sword can be removed from the stone: they're a test of strength: it can be done - but can you?

 

[maxflinn]

One of the reasons it's unwise to claim 'impossibility' for an outcome here relates to the reality of combinations 'in the field'.

Windows OS is a great example. It's not only that it's badly written; it's tackling a genuinely impossible task: catering for a moving target of billions of combinations of hardware and drivers - it will inevitably crash/fail/suffer unresolvable conflicts. I often think about the Challenger shuttle, too: engineering and QC of the highest calibre; O-rings perfectly fit for task . . . until they got cold - which rapidly butterflied into an unpredicted (though not unpredictable) terminal chain reaction.

Every listening room consists of unpredictably related parts; each DAC is differently susceptible to external power and jitter issues; each system is differently vulnerable and resolving; each listener is differently tuned for acuity. And this is borne out by the universal lack of agreement about what works best. How could it be otherwise?

So when there are concrete material differences between cables - and there's no difference in this respect between those carry high and low-frequency / square or sine signals - it's quite possible these will be audible.

Even this thread shows that differences between USB cables are audible to some - 50%, in fact. The more controversial issue is defining 'audible', which is properly the subject of another thread. This is only a listener survey, not an inquest.

All very sciency sounding, Mark, but you've missed out the real important sciency bit, which is that subtle audible differences between digital cables as claimed by reviewers are simply not possible.



They can't exist, it is a fallacy.

 

[merlin]

it can be done - but can you?

Actually as it stands Item, it can't be done - but are you the one who miraculously can?

Of course if you TELL people it can be done, and they find they can't, they are going to feel a little inadequate. They are therefore likely to pretend that they can, or imagine that they have.

 

[ItemAudio]

More misrepresentation.

Very sad to see this day after day.

Look again at the figures. Only 22% claim to have heard differences - with no indication whatsoever as to whether these tests were subject to expectation bias - scientifically the overwhelmingly most probable reason for these people having this view.

Please stop that: it's ruining a thread which has sporadically reached the level of an adult conversation.

After 141 votes the poll states 50% (31 of 62) of those who auditioned multiple USB cables hearing differences between them.

The remaining 79 votes solely indicate expectation bias: 76% expect no difference.

 

[ItemAudio]

Actually as it stands Item, it can't be done - but are you the one who miraculously can?

Of course if you TELL people it can be done, and they find they can't, they are going to feel a little inadequate. They are therefore likely to pretend that they can, or imagine that they have.

As I've said before, I strongly agree with Alan's point that acuity can be selectively trained over time - and that unusually high levels of sensitivity can be acquired to specific stimuli when focused on for sustained periods. That is, after all, how brains work. It's not about 'being a good listener': it's about acclimation.

But on balance I don't believe blind testing is any more useful than sighted. I'm always very impressed when someone passes a blind test because the dice are so heavily loaded against doing so! As I've said many times, I don't have any special skills in that area, and I don't approve of the word 'miracle'.

 

[Alan Sircom]

'House' sound for a USB cable? I need to shake that cable's designer by the hand.

How on earth can a house sound be designed into a USB cable?

I don't know. But reversing this, how can the one person who speaks out about specific brands correctly identify the brands, when they didn't know what brands were in the test? I'm not filtering this - these weren't correct guesses out of six wild guesses and I wasn't asking people for that level of analysis (because I didn't think it was possible). They only spoke out when they recognised specific characteristics in the presentation, and they got it right three times out of three.

The person who commented was the owner of the system and the room, so he was best acclimated to its performance. And he did know one of the products in the line-up (because he supplied it), but didn't know when it was to be presented.

So the fact remains, how is this possible? Sheer chance is a possibility, but randomly naming names - especially at the time names that had no dog in the USB fight - and getting it right is a little uncanny. There are only five possible outcomes:

1. I'm lying
2. There was some cheating going on
3. Bias was leaking in
4. Blind chance
5. There were observable difference

You have to take #1 at face value. Or not. There's no way I can convince you either way on this. Similarly, I tried to prevent the incidence of #2, and I think I was successful. #3 is a distinct possibility, and he could have been cold reading me, but I have a pretty good poker face. #4 is also a distinct possibility, but I asked him for six lottery numbers after the test and I got nothing. And #5 is also a possibility. That's the best I can pull from this.

However, this comes from a listening test to see if there was a point in further listening tests. I'm saying all this because I getting somewhat fed up being told by those who won't do the same test that it's intellectually dishonest to do such a test.