Jump to content
IGNORED

Internet blind comparison of two USB cables


pkane2001

Recommended Posts

This started as a proposal for an internet blind test on another thread: 

 

https://audiophilestyle.com/forums/topic/62140-asr-audio-science-review-forum-youtube-channel/?do=findComment&comment=1113069

 

@PeterSt's Lush^2 cable against a generic USB cable, used to generate a recording that can then be tested by others.

 

Claim:  Lush^2 produces an obviously audible difference that can be detected 10 out of 10 times by Peter and others in a blind test

 

Test: The test will be fairly simple: play music using a PC (iMac) over USB cables, A or B, directly into the DAC. Feed the analog output from the DAC into a quality pro ADC (Apogee Element24, which I've tested and measured before and know it measures well) and record this to a digital file. Rename and randomize the samples and upload them for a listening test. I can even publish measurements for these afterwards, if there's interest. Those hearing obvious differences can use an AB or ABX comparator that produces a digitally signed result and the result can then be shared when reporting your findings (FooBar ABX comparator, for example, also DeltaWave and a few others that I've seen). You're welcome to use any other testing mechanism if you want to listen for yourself, but your results will be harder to be confirmed by others. 

 

Lush^2 cable will be configured as follows, per Peter: A:B-W-Y-R, B:B-W-R
 

Before setting up the test, I'd to solicit any ideas here on how to improve the test and methodology, so please post your thoughts, concerns, and ideas here.

Link to comment
22 minutes ago, sandyk said:

 When will you finally get it , that to hear details such as this ,you can't use any nondescript PC, or worse still a stock standard Mac Mini, which as many members have already verified markedly improves after an Uptone Linear PSU designed by the talented John Swenson is fitted to replace the original SMPS .

 It may be news to you, but it isn't to many 100s of A.S. members that the front end DOES matter !

 

I'll use a battery powered laptop, which I normally use for measurements. There's no AC interference and no ground loops. I just ran a J-Test through the whole chain (PC->DAC->ADC->PC) to see if noise and jitter are a concern. Looks clean to me. Not sure where the -140dBFS spike is coming from at 250Hz but that's tiny anyway. The DAC is Holo Audio Spring v1 (with updated USB).

 

image.thumb.png.28914458c57050b2b2c3a1d9212ad0d7.png

 

image.thumb.png.c9e5f392ee45e672db84f309f200b70a.png

Link to comment
4 minutes ago, davide256 said:

Sure.

 

1) you are limiting/coloring results by the digital capture solution

2) any tester is adding limitations/coloring results by the solution they use to do a second generation decoding

 

We use 2nd hand observation for places we can't go like Mars, but that always flawed compared to 1st hand direct observation

which provides more sensory data.

 

Well, that's easy to verify (and has been done before, just not with this ADC). I'll run multiple iterations of the recording in a loop through the ADC, and if you can hear the difference, then the "coloring" is significant. If you can't, then it's audibly transparent.

Link to comment
12 minutes ago, audiobomber said:

Looking forward to the test.

 

I had an Oyaide Class S (silver) USB cable in-house for a couple of weeks, couldn't stand the overly bright sound in my system. I replaced it with a cheaper Oyaide Class A (copper) cable, which did the trick. I don't understand how someone would not be able to tell these two apart.

 

OTOH, I couldn't hear a difference between the Oyaide Class A and a Curious cable I acquired subsequently. Unfortunately, printer cables, of which I've tried many, don't compete with these, IME.

 

That's what I'm hearing from others. If these differences are so obvious and audible, they shouldn't be hard to capture and to reproduce.

Link to comment
Just now, davide256 said:

You trust your instruments not to miss anything, that what they can’t detect can’t be heard by human ears. Quite a prejudice. The interesting parts of science that we learn from are what we can observe but not yet explain. To do this properly you need to observe directly, not depend on unwise assumptions from unneeded extra steps. Best practices approach would be to take another cable with similar length, similar A-B ends to the Lush 2, wrap them with like covering, label 1 X, the other Y and send to a series of reviewers to serially provide feedback on a standard form. 5 would probably be enough for directionally correct results if you insured all had good systems, much larger sample required if screening crappy hardware was done at the tail end of collecting observations


Sorry, since what I’m proposing is a listening test, I don’t see how your criticism about instruments applies.

Link to comment
8 minutes ago, sandyk said:

 If you are unable to hear the differences directly, as I suspect,  it will almost certainly be a waste of every participant's time by further proceeding as it is highly likely to be due to your equipment simply not being transparent enough.

ray_dude is also saying something similar.


I don’t see how any of this follows from the fact that I can’t hear the difference.

Link to comment
12 hours ago, PeterSt said:

 

Paul, although it was not discussed extensively, nothing of that kind would be the idea. I said (and you did not disagree) something with this gist:

 

a. I provide the music to you.

b. It should be full tracks, me taking responsibility of that and if needed, me distributing your recorded files to free you of the responsibility (you half-agreed and wanted to cut the tracks, maybe).

c. I'll try to find interesting music, despite it won't be the style of most.

 

No Foobar or anything is required, because people would receive an album with a "show" on it (my show). The tracks will have their original name (Artist + Track Name).

Exactly half of the tracks is recorded with the printer cable involved, the other half with that other cable, starting with L.

 

This is how my A/B would work as this is no tiring stupid A/B, but just listening as we always do.

In this case there would be an added feature : There's a special 00- track recorded twice, and of that track you'd indicate which is recorded with which USB cable. You could call that Cable A and Cable B, but I think you should name that to the real names. Thus Printer vs L. That way people could try to find their own Lush cable after first claiming for themselves the recognition in the 00- track. If the recognition exists in their imagination only, then this is fine BUT the same imagination may happen to the test tracks and it would work (OK, could work).

Those not owning a Lush will have a different imagination, but it could be one of "lush sound" (the name stipulates). Whatever their imagination, they should find that back in the test tracks.

 

If you read the above again, you will more and more think it is voodoo. But it really is not and when the difference is in there (captured by your microphones (stereo !), it will be 10 out of 10.

To be clear :

 

People can play the tracks as often as they want but this is not advised because it is meant as a listening evening**, with slightly not your kind of music. 😏 What I would do is listen all in one go per the sequence in the album and once I have the hang of it, when finished repeat the first tracks until the point I got the hang of it.

 

**) See next post.

 

I don't know if you realize it, but you're describing exactly an ABX test. There are two tracks that are properly labeled -- references, and there are two (or more) files that are not labeled. Your job in an ABX test is to match up unknown files to the reference. How you do this (listen to the whole track, listen to 5 seconds at a time, repeat the track multiple times, compare it multiple times to each of the references, etc.) is exactly what ABX test allows you to do. There's no time constraint or requirement to do quick switch-overs. How you match an unknown track to a reference is up to you, as long as you don't cheat ;) For some reason people have an allergic reaction to these three letters (ABX), but it is really just a simple listening test comparing one unknown file to two known ones. For any obvious differences, it should be trivial to pass such a test, using your normal listening methods.

 

ABX tests also have a training mode, where you can spend as much time trying to identify the unknown tracks, until you feel confident. Then, run the real test in exactly the same way (again, as much time as you want, repeat tracks and references as many times as needed, etc.) and record the result.

 

ABX test has the benefit of producing a confirmed test result. Someone claiming I got all 10/10 right has no way to have this result confirmed. With an ABX test, there's a signed document that can be verified. Certainly cheating is still possible, but at least it's 


But anyway, to make the test easier to those who are allergic 😎, I've already said, use the ABX tester if you want your result to be meaningful. Otherwise, use any listening method you prefer, except that the result is not something we can confirm.

 

 

 

 

Link to comment
11 hours ago, PeterSt said:

 

Maybe I must back out on this at first. Because what if this works after all, by means of capturing the analogue output of the DAC. I mean, I propose a listening evening, but I don't really see myself listening "an evening" to hollow sounding unrealistic stuff.

If it does not work (I don't hear a difference anywhere**), we can always use the microphone method).

Btw, an other reason I may now like to go your way, is the awkwardness for you.

 

**) I also said that I should have a pre-listen to two tracks (but make those the same tracks recorded by otherwise unidentified printer and L cable) and if I don't hear a difference, then we might just as well forget it. So I think this step is a necessity, to save you all the trouble for when it won't work out anyway.

 

So do we agree to capture analog outputs for this initial test? I'm not against doing the microphone test after this, although I think it'll introduce way too many new variables into the test, like microphones, speakers, amps, cables, power supplies, etc. that have nothing to do with the USB cable and can all generate obvious differences and distortions. 

Link to comment
6 hours ago, PeterSt said:

Houston ...

 

Paul, the last paragraph in my previous post makes me realize that we need to decide something for the sampling rate. This is a bit unexpected for myself, and somehow does not apply to capturing by microphone. It does after all of course, but:

 

The files you will receive will be 16/44.1. At least that is my thinking at this moment. Now what happens in your situation when you play that from your iMac to your DAC and look at the outputs ? Heck, this is a subject in itself.

How do I arrive at this subject ?

 

Well, my first idea for a general test was to play back a by you recorded file that contains sufficiently broad data (for frequencies) that something should differ with different cables. So a first thing to check - and this can now happen if you indeed record from the outputs NOT by microphone - whether ever something is left from the staging, the depth, the holographic picture, the detail and so much more. This is obviously related to your A/D machinery and my expectations are not good at all. Btw, this is again my reference to e.g. Mani applying similar recordings with whatever he thinks is good enough for A/D, because a first thing what will happen is that all the life is sucked out of it. I know, Paul, you will claim sufficient transparency from your A/D, but it really does not work OUT like that. it is actually the same thing as the AmirM thing. You (or we) overlook giantly on these matters.

 

Your playback chain (!) will most 99.99% probably smear all detail to a mush without life. You won't know that because you are used to it, but relatively to what I am used to, it will be quite unbearable. This really, seriously, makes nothing wrong with your playback chain, but for me it really implies that nothing works out any more. Well, will it really not ?

That is thus my first test. It is necessary. It belongs.

 

Trying to avoid 20 subjects, my actual subject for this post is thus the sampling rate of the recording. This definitely can't be 16/44.1 when your process under way made e.g. 24/352.8 of it. You also most certainly can't sustain the 16/44.1 because you would be operating without reconstruction(-filter) *if* your DAC would allow that in the first place.

So it seems obvious that you'd need to apply the highest sampling rate** your ADC allows for with a bit depth of 24 (if it allows for 32 you should not use that because only a few people would be able to play it).

I now envision your HOLO, and if nothing changed it could be a multibit with possibility of NOS. So would that be better then ?

**): Make that 24/192; I assume everybody can play that back.

 

This is a nice quandary;

If I am provided a 24/192, I can't apply my own filtering any more. And this is crucial (it's do or don't for good sound). If the sound is smeared at playback on your side, the detail is gone forever. Mind you, my examples will be all about detail in various frequencies.

 

And so again, at this moment I don't see how the test could work. Help.

 

Somehow ... somehow I still feel that with amplification of all what happens in there - this is my thesis on playing through loudspeakers and record *that* by microphone - still has that chance.

So I really seem to be back at square one, and I was ignorant to suddenly think otherwise ?

Besides it is clearly on my mind that if that fully developed sound (in the room) is captured at 16/44.1, suddenly all is smeared into something which is completely natural. This is a hunch. A feeling. The quandary in that one is that I probably won't listen for hours to that poor recording (I assume that it will work out like that, Paul, and this is totally unrelated to you or the gear or the room; it will just be poorer (hollow sounding) quality).

 

Peter, I think you're overcomplicating things. The test is not to simulate what you have in your system, with your filters and your sampling rates, but rather what happens in a typical audiophile system. Source material of 24/48k or even 24/192k will work just fine with my equipment.

 

But I have to ask again, how do you think that playing through extra amps and speakers, recorded by a microphone, and then recorded using an ADC will not cut off those higher frequencies above 192kHz? How many people using your cable have a tweeter capable of producing frequencies up to 192KHz? How many of the amps that they have are capable of this? I'd say not that many.

 

Holo Spring is a good one to test with because it can run in NOS mode (no oversampling, no jitter recovery) and in oversampled mode. It can also accept sampling rates up to 384k and, if desired, I can feed it DSD256. But I have other DACs and DDCs I can use for the test if for some reason Spring isn't good enough.

Link to comment
2 hours ago, davide256 said:

 the effect of the Lush 2 is not to change  recorded bits, its to allow the connected gear sending/receiving the bits to behave better, reduce signal contaminants that stress USB I/O circuits, resulting in poorer real time performance. You can't hear this if the Lush 2 is not in circuit. Your methodology

just adds the problem back in.

 

Did I argue about changed bits? You seem to be on a different track than the rest of the discussion. If you see a problem with what's being proposed, please propose a different methodology.

Link to comment
22 minutes ago, PeterSt said:

 

It is about that, Paul. I have no clue why you mention this (again), as it applies nowhere.

I know, I must be thick ...

I probably don't read well enough.

 

You seem to avoid the full tracks. This confuses (it allows you to randomize (story) etc.). You also avoided the microphone recording. So much so, that I even missed it at first.

 

Now I am over complicating things. ? I am not. I think I know what will happen.

Think about you being on the safe side. Because is nothing is audible, who is right ?

 

We'll see with a first test prior to the real stuff. 😇 Yesterday I formed a nice list of music which is bearable for everyone. Next up is the list with the real stuff. This will also contain my pick on this first test. I am going to work on that list right now.

Full tracks. Not stupid small pieces.

 

I only avoid full tracks for internet distribution when there is a copyright consideration. This is the law where I am, and I'd rather not break the law if I can avoid it ;) 

Link to comment
15 minutes ago, davide256 said:

I stated earlier  the valid way to test, a classic A-B blind test.. your test is the equivalent of hearsay... the testers aren't in the room to observe first hand

anything that your test method is not capturing

 

"Best practices approach would be to take another cable with similar length, similar A-B ends to the Lush 2, wrap them with like covering, label 1 X, the other Y and send to a series of reviewers to serially provide feedback on a standard form. 5 would probably be enough for directionally correct results if you insured all had good systems, much larger sample required if screening crappy hardware was done at the tail end of collecting observations "

 

That's a good test, but much harder to conduct properly in the days of COVID and, in general, for a distributed audience spread over multiple continents. Maybe in a year or so we can do something like that. For now, sharing recorded files is the best we can do.

Link to comment
28 minutes ago, PeterSt said:

 

Right. So you do that on your own decision, do not communicate it despite me mentioning it several times, including a solution.

No wonder that I come up with "shifting goal posts".

 

Paul, please, you seem to have your own way and go your own way. Go ahead. But then I am out.

I really don't know what way of working this is but it is not my way. I am really sorry (to be Dutch or something).

 

You seem to be having a conversation with yourself, Peter. I'm just observing :) You're in, then you're out, then you're in again, and now you're out. I've not changed any of the parameters that I originally proposed, except to adjust to yours and others requests, so I don't quite understand this dithering. I even offered to do a mic recording after the analog loopback test. 

 

If you don't want to participate, that's fine. I'll publish what I record and others can decide if they want to listen to it or not, and whether they hear the difference. 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...