[BiggerHead]

... and I'm sorry but in spite of the rules above, Ethos is still extremely important.   Very few people are capable of proper scientific reasoning, and NOBODY has time to go through all of it for everything, and while all those people who are and do might be convinced by the reasoning alone, to an outsider who can't quite grasp the reasoning or has no time, the winner can still be unclear from the logic or illogic alone.   Reputations built up among peer review of reputably impartial people are important.  Without peer review most people cannot know what to trust.  Who is deemed qualified to review, and by whom?  Ethos matters.  Ethos doesn't make the facts right, but it's a critical part of the process of spreading trustworthy information and gaining acceptability of it.  Even in reviewing papers among and by esteemed experts, a reviewer must assume some level competence in the numerous subtle places that mistakes could have been made.  Of course confirmation still matters, but again, confirmation by trusted experts matters more.
 
Even for pyscological/subjective effects, people who have to face actual scrutiny and review often learn to guard their own observations against subjective influence better than people who don't.  Of course on the internet we don't often get a credible sense of Ethos among ourselves for various reasons, but we can get it for referenced works.  
 
In the end, all human knowledge comes down to trusting observations of others, and I'm afraid that simply balancing the volume of trustable observations with non-trustable ones, won't get most people anywhere, because sadly the balance is probably in the wrong direction.

 

[BiggerHead]

@m I'm agreeing with you on the usb rate distraction even if I didn't say it well.

 

[mmerrill99]

@m I'm agreeing with you on the usb rate distraction even if I didn't say it well.

Sure, I know that - I just wanted to clarify the difference between USB throughput & USB framerate & correct those who said I was "misrepresentation of the facts (the speed of the digital audio signal)"
There are so many distractions but very little in the way of cogent counter argument or attempts at refutation of the core points - just personal attacks

 

[mmerrill99]

Actually, I'm wrong & to prevent being accused of "misrepresentation of the facts", I need to post the correction here.

In calculating the bandwidth of USB high speed I forgot to take into account the overhead bits of the protocol. A bulk IN transaction has an overhead of 24 bytes and a payload of 512 bytes. That's a total of 536 bytes. The timeslot between is 7487 bytes wide. Without the need of bit stuffing there is space for 13.968 packets. Having 8000 Micro-Frames per second we can read data with 13 * 512 * 8000 B/s = 53.248 MB/s & not 65.5Mbps

The real-world throughput is more like 30-40Mbps

 

[nick_charles]

Yes, USB bit error rates are not worth worrying about - they are insignificant
Yes but everyone seems to be forgetting the difference in cable shielding between cables which can result in different levels of noise incursion @ different spectrums
I was talking about USB high-speed protocol, which most USB devices currently use which has a theoretical throughput of 480Mbps but an actual real-world throughput of about 65.5Mbps. The USB frame rate is a different thing - for USB high-speed it is 8 frames per mS or one frame every 125uS.Each frame can carry 1024 bytes which gives 65.5Mbps
My initial point was that cable shield differences between cables can certainly make a difference & I posted measurements for USB cable & a technical document explaining Shield Current induced Noise

BTW, thanks for trying to introduce some rationality to this thread!

 
 
A few questions.
 
1. When transmitting 16/44.1 or say 24/96 audio data what frequency does the USB link use to send data ?
2. The Rane paper describes issues with 125ft lengths of cable which as I understand it way longer than USB can support ?
 
In the graph above the horizontal axis does not start until 30Mhz so low speed transfers are completely missed.?

 

[BiggerHead]

None the less you haven't answered a couple of my points.  Induced noise in a ground loop comes in through the whole loop, not just the cable.  This includes the two power cables involved and the outlet and/or wall wiring between them.  
 
But lets just say in some configuration of all that some noise can be controlled by shielding of the cable as in your plot, has anyone shown how this reduces IMD in acutal DAC outpu signals?  Surely it's not hard to show.  There should be some pure tone that one can play and capture back, and see an audible distortion signal and watch that change with cable configuration.  A lack of that evidence doesn't strictly mean it isn't true, but well given what these cables costs, it kind of does mean it isn't true.
 
So does it exist?

 

[BiggerHead]

   
 
A few questions.
 
1. When transmitting 16/44.1 or say 24/96 audio data what frequency does the USB link use to send data ?
2. The Rane paper describes issues with 125ft lengths of cable which as I understand it way longer than USB can support ?
 
In the graph above the horizontal axis does not start until 30Mhz so low speed transfers are completely missed.?

 
He seems to have agreed that bit errors are not a problem.  I was confused on his stance on that briefly, but I think data transfer quality is not in debate.
 
Hmm.. maybe that's not the point of your questions.

 

[nick_charles]

   
He seems to have agreed that bit errors are not a problem.  I was confused on his stance on that briefly, but I think data transfer quality is not in debate.

 
But the effect of noise differs with transfer speed (frequency) according to the graph and low speed (audio)  would be off the graph (as I understand it) so the impact of noise is unknown

 

[mmerrill99]

Yes, USB bit error rates are not worth worrying about - they are insignificant

Yes but everyone seems to be forgetting the difference in cable shielding between cables which can result in different levels of noise incursion @ different spectrums

I was talking about USB high-speed protocol, which most USB devices currently use which has a theoretical throughput of 480Mbps but an actual real-world throughput of about 65.5Mbps. The USB frame rate is a different thing - for USB high-speed it is 8 frames per mS or one frame every 125uS.Each frame can carry 1024 bytes which gives 65.5Mbps

My initial point was that cable shield differences between cables can certainly make a difference width: 350px; height: 217px">
[/URL]

A few questions.

1. When transmitting 16/44.1 or say 24/96 audio data what frequency does the USB link use to send data ?

USB devices negotiate with one another at initialisation to find the highest speed that they both operate at. Thereafter the USB link operates at this speed - doesn't matter if 16/44 or 24/96 data is being transmitted

2. The Rane paper describes issues with 125ft lengths of cable which as I understand it way longer than USB can support ?

Yes but the paper is about shield current induced noise - not just USB cables. The Rane paper actually mentions using four cable lengths in their measurements -- 125 ft, 50 ft, 25 ft, and 10 ft.

The ferrite USB cable measurement is a direct measurement of RF noise differences between two USB cables

In the graph above the horizontal axis does not start until 30Mhz so low speed transfers are completely missed.?

I'm not sure I follow your question - the study is about RF noise

 

[BiggerHead]

ok, 

   
But the effect of noise differs with transfer speed (frequency) according to the graph and low speed (audio)  would be off the graph (as I understand it) so the impact of noise is unknown

but what "effect" are you getting at?  error effects?  We're not concerned about the effect of the noise on the data transfer.  We're concerned about the effect of noise on the DAC, and maybe if relevant then the effect of transfer speed on the noise.

 

[mmerrill99]

None the less you haven't answered a couple of my points.  Induced noise in a ground loop comes in through the whole loop, not just the cable.  This includes the two power cables involved and the outlet and/or wall wiring between them.

But lets just say in some configuration of all that some noise can be controlled by shielding of the cable as in your plot, has anyone shown how this reduces IMD in acutal DAC outpu signals?  Surely it's not hard to show.  There should be some pure tone that one can play and capture back, and see an audible distortion signal and watch that change with cable configuration.  A lack of that evidence doesn't strictly mean it isn't true, but well given what these cables costs, it kind of does mean it isn't true.

So does it exist?

I haven't seen any such plots but I suspect that we are talking about some difficult to measure distortions such as noise floor modulation that is correlated with signal

 

[BiggerHead]

ok, so you're saying the noise of concern is 100% outside the audible range, to one part in 10,000, this could really be true for some sources,  that it's not filtered or referenced out, and that the IMD effect is always just below or at the audible noise floor.  Surely it would be possible to make a test environment where this could be increased by 10x to bring it above the noise floor, if it's real. This is really starting to sound like chasing a white rabbit.  It's always just around the corner.   

 

[mmerrill99]

 

He seems to have agreed that bit errors are not a problem.  I was confused on his stance on that briefly, but I think data transfer quality is not in debate.

But the effect of noise differs with transfer speed (frequency) according to the graph and low speed (audio)  would be off the graph (as I understand it) so the impact of noise is unknown

I think there is some confusion here - USB uses packetised delivery protocol - there is a frame every 1mS & if high-speed is the negotiated speed being used there is a further sub-division of a frame into 8 subframes i.e. a subframe every 125uS

 

[BiggerHead]

Anyway, if the worst the effect gets is the level of the noise floor of the best DACs that one could try to use to measure it, or even also at the noise floor or worse dacs that let it through more, then I'm pretty ok with it and pretty confident it won't affect my music quality.

 

[mmerrill99]

ok, so you're saying the noise of concern is 100% outside the audible range, to one part in 10,000, this could really be true for some sources,  that it's not filtered or referenced out, and that the IMD effect is always just below or at the audible noise floor.  Surely it would be possible to make a test environment where this could be increased by 10x to bring it above the noise floor, if it's real. This is really starting to sound like chasing a white rabbit.  It's always just around the corner.   

My hypothesis is that noise floor modulation is at the heart of this issue & it is caused by IMD of RF noise. Now RF noise only produces IMD when there is a signal present & it is a difficult thing to measure noise floor modulation in the presence of a fluctuating signal.

When I used the ferrite compound USB cable I heard a very similar change in the sound as when I used an Intona USB isolator with a standard USB cable. Hence my hypothesis