Digital Cables and Noise

image credit: ECN Magazine

The notion that a digital cable, including Ethernet and USB, can be susceptible to noise is often met with harsh criticism. If we then add fuel to this fire by further suggesting that this noise can cause an audible difference in our connected hi-fi, we are typically met with derision [footnote 1]. Some people contend that for all intents and purposes, digital systems are noise-proof, the proof being bits are bits. Therefore there's no way a digital cable can possibly make any difference when it comes to the sound coming out of our speakers. The fact that noise in digital systems including digital cables is a well-known problem outside the world of hi-fi doesn't appear to make one bit of difference. Let's see if we can separate the signal from the noise.

Let's start with some facts about Ethernet and USB cables. Here's some very good general information from an article in ECN Magazine titled, "Differential signaling best practices" written by Clark Kinnaird, Industrial Interface Systems Engineer, Texas Instruments:

"Differential signaling is used for noise immunity in Ethernet, RS-485, CAN and USB. In ideal cases, all common-mode noise is rejected. In real-world applications, there are several design techniques and component parameters to consider in order to keep the data flowing with high confidence. Differential signaling is used in most interfaces, which sends digital information over cables. Although requiring two signal wires rather than one, differential signals are much more immune to noise than single-ended signaling.

"The basics of differential signaling are well-known, taking advantage of the noise rejection which affects both signal wires equally. This is illustrated in Figure 1, where a balanced differential signal is transmitted on two twisted signal wires (twisted-pair).

"Popular electrical standards such as USB, Ethernet, RS-485 and CAN use differential signaling and balanced twisted-pair media to provide reliable high-speed communication.

"In practice, designers should keep in mind that no real system has the ideal performance of a theoretical model. There are several key sources of errors and noise that should be considered."

"In practice, designers should keep in mind that no real system has the ideal performance of a theoretical model."

The key sources of noise referred to in this paper include line-to-line impedance imbalance, transmission line length, reduced immunity during idle and cross-over times, and hysteresis.

image credit: ECN Magazine

Let's dig into this a little further. From "A Ham's Guide to RFI, Ferrites, Baluns, and Audio Interfacing" by Jim Brown ("The basis of this tutorial is a combination of my engineering education, 55 years in ham radio, my work as vice-chair of the AES Standards Committee working group on EMC, and extensive research on RFI in the pro audio world where I’ve made my living"):

"Several cable defects (essentially manufacturing tolerances) certainly can and do convert this 'common mode' antenna current to a differential signal (that is, a voltage between the signal conductors), but that is rarely the most powerful coupling mechanism. One common defect that affects both balanced and unbalanced cables is imperfect construction of cable shields. In even the best 'real world' balanced twisted pair cables, there are imbalances in the capacitance between "red" and "black" conductors to the shield on the order of 5%. [B. Whitlock, JAES, June 1995] In balanced paired cables that use 'foil/drain' shields, there is even more imbalance in the inductive coupling between each conductor and the shield. Noise (or RFI) coupled by this mechanism is called 'shield-current-induced noise,' or SCIN. [N. Muncy, JAES, June 1995] All three of these mechanisms convert shield current to a differential signal at system input and output terminals."
Essentially both of these papers address how to stop noise from coupling to a digital cable and passing this noise along to the connected system.

image credit: Texas Instruments

This paper from Texas Instruments, "A Survey of Common-Mode Noise", offers measured results of radiated noise being induced in Cat 5 UTP cables (Beldon part #1588A) of varying lengths due to their proximity to various noise emitting devices. It also offers a nice introduction to common noise sources including EMI, RFI, and Ground Loops in digital cables. Here's their summary:

"It appears from Figures 2 through 12 that cable length made very little difference on the induction of noise in the described environments. There was some noticeable attenuation on the longer cables that could be attributed to the dc resistance of the cable and normal line loss factors. This was not always the case when measuring the drill motor EMI. The longer the cable, the better chance that high EMI fields can couple to it. It should be pointed out that if the same test were repeated, the results would be different because the noise level varies constantly. The noise was very unpredictable in all cases. We will never be able to eliminate all the noise in our environment."
Here's another Ethernet example from the same Jim Brown paper, "A Ham's Guide to RFI, Ferrites, Baluns, and Audio Interfacing", but this time he's talking about his pro audio experience:
"The author has demonstrated that sensitive microphone circuits connected with unshielded CAT6 cables are actually less susceptible to noise pickup than shielded balanced pairs, both at audio frequencies and at VHF/UHF!

"This is true because the degradation in rejection due to imperfections in the cable tend to be greater than the relatively small benefit of the electric shield!"

To say that noise is not an issue in Ethernet cables is to simply ignore some very basic facts. The question then becomes—can this noise really affect our analog systems?

Here's Ham radio user Joel Harrison talking about chasing down the source of noise in his system in "Slaying the Grim Reaper of Internet Router Noise on 6 Meters" (my emphasis):

"Locating the source of the interference can be relatively easy in this case. With your radio on and listening to or seeing the interference on a panadapter, turn off (remove power from) your internet equipment. If it is the source the birdies will go away. Mine did, but I also noticed a problem I had not previously identified at the time. My noise floor dropped 10 dB! I had attributed my increased noise floor on 6 meters over the years to a variety of problems, from line noise to consumer devices in my neighbor’s homes, but definitely not something in MY home!

"When I powered up the internet modem and routers sure enough the birdies were back and the noise floor had increased."

Here we have a case where a router and/or switch radiated RF which coupled to an Ethernet cable adding 10 dB of noise to the connected analog signal as well as spurious frequency related noise. The solution to this problem involved using ferrite cores on the Ethernet cable.

What about USB?
From microHAM's support page titled "RFI":

"The problem: microHAM have been receiving reports of distorted microphone audio in otherwise normally functioning micro KEYER II. Although micro KEYER II has been tested for operation in high RF fields, we have found that when the antenna is very close to the ham shack, or antenna is not properly balanced (lack of effective balun), or open wire feedlines are used, radiated RF can couple to the shield of the USB cable and from the shield of the USB cable into the microphone preamplifier resulting in distortion."
In this instance we have radiated RF entering and adding distortion to the attached analog system through a USB cable.

Here's some additional information from outside the RF ham world, "High Speed Differential Interfaces and the Need for Common Mode Filtering and Protection" Jonathan Harris, ON Semiconductor, talking about Hi-Speed USB 2.0:

"In the operating environment seen by electronics today there are numerous sources of electromagnetic interference (EMI) and radio frequency interference (RFI). This is due in large part to the increased use of RF technology. These types of interference result in the need for common mode filtering in applications utilizing differential interfaces. Although the desire in utilizing differential signaling is to minimize the effects of EMI/RFI, the effects are not eliminated entirely. The differential signal can become corrupted by external noise such that it is not discernible at the receiver. Also, with the noise having coupled into the electronics inside a device, other circuitry where differential signaling is not incorporated may be affected and cause addition problems."
Of course some people within the hi-fi world also talk about USB cables and noise. Here's Robert Watts the designer of the Chord Hugo, among other well regarded DACs, from a post on Head-Fi talking about USB cables and noise:
"...RF noise can leach in through the ground, then you have coupling through the pins, then you have the RF noise that the receiver creates due to noisy inputs. A lot of so-called audiophile cables actually are RF poor and create more noise.

"Some people actually like the sound of noise floor modulation, because it sounds brighter and superficially more detailed. So we have a lot of variables, which is why there are so many opinions.

"Myself - I used to be very bothered about cables, I am very much more relaxed about it now. Make sure your digital connections have good RF performance, then Hugo will sound fantastic, simple as that."

"A cable is a source of potential trouble connecting two other sources of potential trouble." a joke among electronic systems engineers

Loose Ends
If you read through the papers I've linked to, you'll find an array of real-world problems as well as real-world solutions pertaining to noise and digital cables. Some people recommend using unshielded Ethernet cables, including well-regarded specialists like Dr. Howard Johnson (if you'd like to read another example of experts differing opinions on how to deal with noise in Ethernet, read this), while Texas Instruments and others suggest "Use high quality CAT5E or better cable in implementing network systems. If possible use shielded cable."

If we want to talk about how noise from an Ethernet cable that's connected to a computer which in turn is connected to a DAC via USB can impact the analog signal, we have to talk about things like current loops which is a subject well beyond my knowledge base. Someone with the requisite knowledge and experience has offered a number of papers on this subject which I've included below. Here's something to get us going from "Understanding, Finding, & Eliminating Ground Loops" by Bill Whitlock, President & Chief Engineer, Jensen Transformers, Inc.:

"Any signal accumulates noise as it flows through the equipment and cables in a system. Once noise contaminates a signal, it's essentially impossible to remove it without altering or degrading the original signal. For this reason, no system can be quieter than its noisiest link. Noise and interference must be prevented along the entire signal path. Delivering a signal from one box to another may seem simple, but when it comes to noise, the signal interface is usually the danger zone, not the equipment’s internal signal processing."

Clearly I've only skimmed the surface of this issue and may have raised more questions than I've answered. I hope that's not the case. The simple takeaway I'd like to highlight is the fact that digital cables are not immune to noise and this noise can directly impact the quality of an analog signal in an attached analog device.

" cables are not immune to noise and this noise can directly impact the quality of an analog signal in an attached analog device."

It also would have been nice and tidy to have arrived at some specific sound advice regarding digital cables, noise, and hi-fi. Seeing as our EMI and RF environments vary as does our gear's ability to deal with it, not to mention our listening preferences and proclivities, there is no one single solution to what is a multifaceted site-specific problem.

One approach is to just shrug this information off based on the belief that EMI and RF noise levels in our home and in our network-connected hi-fi are just not worth worrying about. There's no experience like no experience. Another approach is to try different solutions including different digital cables and see if they make a difference. There's no experience like experience.

Footnote 1. The last time around this topic in my article "Cables, Bits, and Noise..." I made a mistake. When describing an incident involving noise entering my ICs from a Rega 'table from an adjacent USB drive, I said that the noise traveled over the ICs, into an A/D converter, over FireWire, and into my MacBook Pro where I saw a roughly 20 dB increase in the noise floor. This noise level was present even with no signal being passed.

As some readers correctly pointed out, what actually happened was the noise from the USB drive became part of the analog signal once it got into the A/D converter so the noise level I was seeing in my MacBook was due to the noise in the signal being passed through the A/D converter and through the FireWire cable. My apologies for overlooking this rather obvious explanation. This clarification does not invalidate any of the recommendations I made regarding eliminating this noise, and of greater importance, people have had nearly identical experiences with digital cables as detailed here.

Recommended Reading on Current Loops

Mesh And Loop Current Methods
Understanding, Finding, & Eliminating Ground Loops
Digital Logic Current Flow
Breaking Ground Loops with Functional Isolation to Reduce Data Transmission Errors
Understanding And Controlling Common-Mode Emissions In High-Power Electronics
Ground Loops: The Rest of the Story

CG's picture

"In theory there is no difference between theory and practice; in practice there is." - Lawrence P. Berra

Applies here, too...

junker's picture

The Audiophile Industrial Complex keeps trying to run with one...

TCP/IP is packet-based with a header checksum and resend capability. USB Audio it is not. Much superior and an incredibly more robust protocol. $5 shield CAT7 >> $1000 Gucci CAT 7 cable.

Michael Lavorgna's picture
...etc has absolutely nothing to do with the noise described here. I referenced a number of papers which explain why this is the case. I also provided real-world examples of how noise can couple to an Ethernet cable. So the protocol is also irrelevant since we are not talking about the effects of noise on data transmission - we are talking about noise coupling to a digital cable and being passed on to the attached device.
"$5 shield CAT7 >> $1000 Gucci CAT 7 cable."
Please point to where I've ever recommended a "$1000 Gucci CAT 7 cable."
junker's picture

While playing music from a NAS over Amazon CAT7 press the pause button on you "memory play" enabled playback software, turn up your PASS INT-30A all the way, and let us know if you hear any noise.

I wasn't quoting you. I am speaking to my experience - I run CAT7 and am more than pleased with it. Thank you.

Michael Lavorgna's picture
...for one aspect of the noise we're discussing. Unfortunately it also misses other manifestations of the same problem. If you look at the ham radio guys mentioned here, you'll see that noise can also affect specific frequencies so unless you are listening to audio content you won't hear it.

I have a question for you - if as you suggest "TCP/IP is packet-based with a header checksum and resend capability.", why do you use Cat 7 cable?

junker's picture

Hi Michael:

A few reasons... one is I am less concerned about noise affecting the signal than I am about the cable emitting noise directly to other wiring (analog, power, etc.). I also was able to buy a few lengths to get just the fit from my wall outlet. Nice and clean. Also, I was able to easily find white cable which look nice in my place. And also I like the nice substantial build quality with a gold-plated metal conductor. For the price on these the difference in price vs. 5e or 6 is a non-issue. Many of these ship Amazon Prime. If I was running long lengths rather than short connector cables I would be okay with CAT6 if it was well hidden, of course. ;-)

Michael Lavorgna's picture won't have to worry as much about your music getting a sun tan ;-)
junker's picture

True that! Normal I like my wiring to be minimal and not stand out. One clear exception is your beautiful green, cloth-sheathed A23 speaker cable. Finances permitting that is cable I would certainly paid one large for!

Michael Lavorgna's picture
...for hi-fi ;-)
VK's picture

I'm happy with the sound of my system right now, and i'm planning some upgrades till the end of the year, but i will not chase possible ghosts in my system. Even experts like Dr. Howard Johnson and Texas Instruments doesn't agree with the use of shielded cables on network systems.

It seems like noise in digital cables will give some headache to researchers for some good years!

Until then, i'll keep enjoying my music collection!

Good article!

Best regards!

Michael Lavorgna's picture
That's really all that matters. Enjoy!
Garth Marenghi's picture

I am sure that there are many possible implementations of digital audio that are susceptible to ground loops and coupling of noise via shields etc. but it is not a *fundamental* problem. It is just another aspect of audio where, with a little rigour and the spending of a few dollars on commodity hardware (as opposed to $1000 on a faux hi-tech cable), the problem, if there is one, can be made to disappear. The buzz-phrase is *electrical isolation*.

Michael Lavorgna's picture recognizing there is one. To say that there has been resistance to the basic notion that digital cables are susceptible to noise would be a gross understatement.

It's also not a question of money, it's a question of identifying the source of the problem, if there is one, and determining the most effective way to mitigate it. This could be as simple as moving things around and apart, i.e. free, to applying ferrite cores, using shielded cables or unshielded cables, etc, etc. To suggest that all it takes is "a little rigour and the spending of a few dollars on commodity hardware" is to oversimplify things to the point of being not very helpful.

And who is talking about "$1000...faux hi-tech cable" besides you and "junker"?

junker's picture

I like my name in quotes...

Has some sort of ominous gravitas to it.

Michael Lavorgna's picture
...I'll remove the quotes.
Garth Marenghi's picture

Yes, we are where we are. I didn't mean that the end users could, or should, have to fix problems themselves (which yes, would be unhelpful), but that the designers of the equipment should. We seem to be in a situation where a concept that was designed to be immune to analogue influences is now assumed to be a minefield of injected noise, ground loops and jitter. In practice some of these assumptions will turn out to be unfounded, and based on advertising hype from cable manufacturers. In other cases the effects will be so tiny as to be negligible. But if a case can be demonstrated where there is a significant problem, then trying to fix it with 'higher quality' cables is just applying a sticking plaster to a problem that shouldn't exist. Maybe the user would be better advised to purchase an isolation interface (as mentioned in a comment above?) to completely fix the problem, but really it would be a shame that this was necessary at all and not built into the equipment to start with (which was the "few dollars" I referred to earlier). Digital audio can work perfectly well using optical fibres or WiFi, so it seems ridiculous that high end audio users are having to worry (or thinking they have to worry) about digital cables.

Michael Lavorgna's picture
Whether or not this presents a problem worth dealing with when talking about hi-fi is up to the individual.
stevebythebay's picture

between the physical medium that carries the bits and the digital bits themselves. Getting noise out of our systems is quite often missed as we focus our attention on taking what we get from sources and delivering the analogue or digital stream into the devices that ultimately play the music. In a perfect world the electric company would deliver an unadulterated wave form with absolutely no other crap to our systems and electro magnetic and other garbage would not interfere with moving power along the way to our various devices.

rtrt's picture

Its late here so didn't read the article beyond the first paragraph sorry.

Surely we can agree that digital data transmission is designed to be very tolerant of noise and hopefully also agree that over the short distances involved in home audio it is effectively a non problem.

Where noise in digital systems can be a problem is when it breaks through into the analogue domain.

I have a great example of that here at home.

I use a laptop - usb cable - dac - amp - speakers, or I should say I did, because there was a level of noise imposed on the analogue outs from the dac that was unacceptable.

I purchased a USB filter, designed for some industrial purpose (c$50) and inserted it in between the laptop and DAC by using 2 usb cables, that removed the noise.

I'm sure there are many more examples, but as I say it's late, so I'll just simply finish by saying for me to buy into (literally and figuratively) an 'enhanced' digital transmission method I have to be persuaded that it will address somehow, some kind of noise that impacts the analogue domain.


Ok not quite finished :)

Of course there is an obvious comment here, which is that if my laptop didn't generate the noise in the first place, or if my DAC somehow wasn't susceptible to it, then i wouldn't need the $50 filter!

Note the laptop wasn't purchased for audio, so it's not fair to level criticism there. The DAC I guess is another matter.

Michael Lavorgna's picture read beyond the first paragraph, I think you'll find that we are talking about similar things.
rtrt's picture

Yep sorry Michael - long day yesterday and similar today.

Wavelength's picture


Actually digital transmission is designed for a price. I designed 5 mother boards in the past. I had 2 accountants strapped to my desk on each of the products.

Not all filters are going to give you what you need.


rtrt's picture

Sure Gordon I was simply citing an example that I had direct experience of to back up my point.

I wasn't suggesting that the specific filter was appropriate for all circumstances, or indeed that a filter would always needed at all.

In my specific case I must get around to substituting a different PC - laptop/desktop to see what if anything changes by doing that.

Sorry -late here again time for bed!

UpTone Audio's picture

For both USB and Ethernet there are complex and sensitive PHY chips which are the very first thing the signal encounters when it enters a DAC or other device. John Swenson has written extensively about how poor signal integrity (noise, impedance mismatch, etc.) causes the PHY and USB processor chips to work harder to decode the bitstream. That extra activity causes "packet noise" (8kHz are for USB2.0, but also broadband) which affects both the ground plane in the DAC and the clocks (despite DAC designers' best efforts with isolators after the USB input processors). A bit similar to what has in the past been referred to as LIM (logic induced modulation).
[John speaks specifically about it in this post:

As your article points out Michael, some of this noise is quite pernicious, but it is important that we look further to understand how exactly it manages to penetrate deep into our DACs, ultimately affecting the power supplies and master clocks.

Michael Lavorgna's picture
As I said, I only skimmed the surface of this issue. People like John Swenson offer much more insight into the complexity and nature of how this noise can influence what we hear.

Here's the first part of our three part Q&A with John which is certainly worth a read.

Wavelength's picture


Thanks for advertising your product on both sites.

I took a look at the circuit board that you are making and your missing a few very key components and the choice of capacitors seems a little off.

You should really say who you are and that it is your product. That is not really what any of these forums are for.


UpTone Audio's picture

Hi Gordon,
Which part of my post regarding taking a look at just how computer and cable variation causes variance in DAC input stages and beyond did you consider an advertisement?
Was it the link to John Swenson's detailed post about PHYs, packet-noise, and signal integrity--which happens to be in the middle of a 600 post thread centered around our product?

Other than my lack of a company signature line (for which there is no provision) here at AudioStream, I don't believe I have ever violated the terms-of-service of any forum I participate on (mostly CA; here just once in a great while). I am sure if I did, the moderators (Michael and Chris respectively) would speak up. Maybe I should do like you and make my avatar a photo of my product. But that seems more like advertising to me.

Plus John Swenson and I have contributed numerous reports and helped lots of people on the forums--having nothing to do with any products we currently make or are developing.
Also, are you suggesting that these audio forums are not for discussion of audio components? That would be news to lots of folks.

Lastly, what's with the glib swipe at the design of out product? I can assure you that we are not "missing a few very key components" and that our choice of capacitors is not "off." Have you read anything in the 200+ post "REGEN Listening Impressions" thread other than raves about its tremendous sonic effectiveness? 250 units sold in 19 days and with zero hype from me.

Best regards,

Alex Crespi

Wavelength's picture


Here's the deal. You use your personal ID and make a post that purely advertising for a product that you make.

If you make posts like that you should always use your Companies ID like you are here.

That is all I am saying. This is not a forum for advertising. You can pay Stereophile/AudioStream for the that luxury, same with Computer Audiophile.


junker's picture

Hello Gordon:

Should we also, in the spirit of fair disclosure, also mention that you are closely affiliated with AudioQuest?

Wavelength's picture


But I am not pushing product.

Michael Lavorgna's picture
...the manufacturer of the Regen, I'm going to ask that you identify yourself as such by adding your name and company info in a signature line attached to all of your comments.


UpTone Audio's picture

Hi Michael,
I am indeed the manufacturer of the newly popular, John Swenson-designed USB REGEN.
I have in the past--and again just now--looked in the member account settings here at AudioStream for a place to add a signature line to identify as a member of the trade. There does not seem to be any such facility to do so. (Nor does it appear that one can edit posts in the few minutes after posting as is the born on many other forums.)
Do you wish to change my user name to UpTone Audio, or do I just need to remember to put my name and company at the bottom of all posts?

Thanks and regards,

Alex Crespi

Michael Lavorgna's picture
And yes, I would appreciate it if you would include your signature as you've done here in any future comments. Thanks!
UpTone Audio's picture

thanks Michael!

Alex Crespi

dallasjustice's picture cables are not immune to noise and this noise can directly impact the quality of an analog signal in an attached analog device."
I hear music in analog. When I see a legit measurement from a DAC's analog output demonstrating differences among various ethernet cables or USB with a proper interface, I'll pay attention. Until then: "To me it's all just mental masturbation." Sammy Hagar.

It's a totally different game if we are talking about s/pdif or AES.

VK's picture

...measurement from a DAC's analog output demonstrating differences among various ethernet cables or USB with a proper interface, I'll pay attention."

Couldn't agree more!

Michael Lavorgna's picture
So you "hear music in analog" yet you are going to wait until you "see" some measurements to inform what you hear. I wonder what Sammy Hagar would have to say about that ;-)
2_channel_ears's picture

Great article, will have to delve into some of the links

Issues of noise take me back to my intern days when I helped another student debug a circuit. There was a glitch that she couldn't figure it out. I found the source and came up with a fix. She couldn't believe what she was seeing and my only explanation was that all digital is not square (when noise gets into the picture).

Michael Lavorgna's picture
And thank you for sharing that story.
kirkmc's picture

If I understand this correctly, what you're saying is that there is _no noise_ in the digital audio stream (the packets, which are later converted), but there can be noise in the line, which is picked up by audio hardware? So when audiophiles rave about how much more "soundstage" there is in expensive cables, or how it is more "lifelike," they're just making it up. But if they have an audio device that allows that noise to get through, then they'll hear something (most likely a single frequency hum)?

Michael Lavorgna's picture
We are not talking about just "single frequency hum." I address this directly above. The easiest example being the Joel Harrison paper.

And no, we are certainly not making things up when talking about the differences in sound offered by different cables.

KBK's picture

I guess lots of folks don't understand the intricacies of how an audiostream is created in the digital domain, with regard to methods of transport and termination, in the situation of a transmission system involving cables, and the internal accompanying termination hardware of the given ADC or DAC.

That they don't understand the fundamentals of the whole package, in-situ. They tend to bring a stated full answer to their partial understanding, which leaves them with an inability to understand jitter generation in these specific cases. Situations and designs that are not designed around how computer based systems (pure non audio hardware signal transfer systems) work.

In essence, a partial understanding of one systems does not bring a full understanding of the potential issues of another system that operates differently, in hardware terms. a system with different requirements in realization of residual noise sources and now that noise/distortion affects the I/O of the given system.

When we stack that upon the 'not well received' factual point that some people's hearing is different, that some people's hearing is more capable than others (a hardcore fact identical in scope as is physical prowess differences and intellectual prowess differences), we get to a real problem that is generally tied to a mental projection.

Realization of all that, as a set, is what compounds the lack of understanding of the hardware issues. It's not just getting past the one thing, it is set of issues that can be compounded to be outside of the realm of understanding, period. The requisite chain of logic to be met, is potent and varied and not just intellectual, but physical, with respect to actual potential reasoning impediment sources.

Thus the screeching of disagreement can be almost fundamentalist in scope and projection.

In other words, a typical audio discussion.

Wavelength's picture


Thanks for opening the can of worms.

All, noise isn't something you can hear. It can effect things in various ways. In digital systems it can throw off say a DAC chip from performing it's magic. But there are other areas as well, such as the I2S link between the data receiver and the dac chip. Noise can effect the amount of jitter an oscillator can have.

Years ago I was perplexed because different audio playback software sounded different. With the analyzers on hand I could see that each one was sending bit true output to the dac, but there still was a difference.

Over a year ago I invested in a USB Compliance test set from Tektronix. This has really opened my eyes to areas I had not seen before. Now I can see differences in cables and host ports and even the effect that endpoints (DACS) have. This has been instrumental in my interpretation of new USB technology.

As some of you know I use to design routers and bridges in my last real job. I designed some of the first 100M Ethernet adapters and so forth. TCP/IP is not that great of a streaming protocol. I do agree that it is error correcting, but if you have to fix the errors then you will hear more than distortion in pops and clicks.

Nothing is perfect, even though you can get about 68MBPS from USB High Speed, you could probably get that much out of 1000M Ethernet, though you can barely get enough for Audio over a 100M Ethernet line.

There is going to be some big break throughs in USB and Ethernet this year and allot of it is going to be the reduction of noise.


Michael Lavorgna's picture
And thank you for opening it further ;-)
There is going to be some big break throughs in USB and Ethernet this year and allot of it is going to be the reduction of noise.
I look forward to getting my hands and ears on some of these breakthroughs.
CG's picture

Put another way, noise is not limited to that hissing sound coming out of one's tweeter. It's an electrical challenge.

Dick James's picture

The reason some expperts recommend cable shields and some don't is because they don't know how the receiving and transmitting equipment terminate the shields. If done correctly, shields are good. If terminated incorrectly, shields can make things worse. Usually what happens is the shield is completely floating or one end is terminated and not both ends. A properly terminated shield will also keep the cable from radiating and receiving noise.

CG's picture

"A properly terminated shield will also keep the cable from radiating and receiving noise."

Under some conditions. It depends upon the frequency due to skin depth. It also depends on the construction of the shield. A shield also may not be completely symmetrical around the contained twisted pairs, which gives a variation in capacitance between the conductors and ground, which in turn can degrade the good effects that twisted pairs provide with regard to coupling.

In other words, it all depends...

Dick James's picture

...but the skin depth is not a problem above roughly 10 MHz. Below that frequency, it takes a very long cable to induce or radiate noise.

CG's picture

I half agree.

The skin depth is not a problem above 10 MHz.

My own measurements - admittedly not on audio equipment - has shown crosstalk issues even with short parallel conductors below a MHz. Sadly...

Dick James's picture

What you stating is an example of conducted emissions and not radiated emissions. The problem is similar in that there is unwanted noise, but the solutions are different. I worked in telecom product design for many years and the solutions to noise problems are not trivial. I suppose that's why there's an industry that can market many potential solutions to audiophiles. Who has the equipment at home to pinpoint the problem? Very few, so trial and error work best.

CG's picture

My measurements were for cables bundled together, as well as traces on different PC board layers with full ground plane layers between. I'd consider that radiated, since there was no Ohmic contact between them.

But, that's semantics pretty much.

Your overall point is one I agree with. There's really not much in the way of standards for most consumer electronics equipment when you get right down to it. Between that and the test equipment requirement, getting things right is usually a trial and error proposition at home. I'm certain that's part of the appeal of headphone listening - way fewer variables, so there's a greater chance for success.

Dick James's picture

...are more common than people realize with most of the noise coming from switching power supplies. One way to rule out a conducted-emission problem is to put a ferrite around the cable and see if the noise is reduced.

Wavelength's picture


Some of the best USB cables I tested did no shield the data lines. They shielded the VBUS/GND pair, but not the data lines. Because in allot of cases the shield will add capacitance to the data lines and slow down the waveform.


Dick James's picture

...state that the differential impedance is 90 ± 9 ohm and the commom-mode impedance is 30 ± 10 ohm (I might be a little off for the tolerances, but the point is the same). The cables have to meet those impedances with or without shield depending on how the cable is sold. The signal getting slower or faster with or without the shield just means a different dielectric was used to meet the impedance specifications. The results are the same regardless as long as the BER (bit error ratio) is 0, or something is wrong with the interface, not just the cable.

24bitbob's picture


Sincere thanks for an excellent and thought provoking article, it's great to see objective engineering considerations being introduced to this topic.

I read your article, and I read through a lot (but not all) of the referenced material. Two main things I picked up from the references: 1) noise is a problem, and 2) the solution to noise lies in the device.

Just one example; in the paper you reference by Nikolas Ledoux - 'Breaking Ground Loops with Functional Isolation to Reduce Data Transmission Errors', he concludes by saying "A hardware simulation of a ground loop provided a real-world example of how ground loops can affect a USB interface, and how a USB isolator, the ADuM3160, remedied the situation".

In all that I read, I still did not see anything that would explain why, for example, an Audioquest USB cable would sound different from one made by Chord, or indeed one made by Belkin.

As you state, we have a long way to go in our understanding of these things, but what I take from this, is that for digital, there is much more to be gained by considering the 'device' rather than the cables that connect them.

Call me a digital cable skeptic, not a cynic. Meanwhile your music recommendations are a must read for me, and for that I add further thanks.


Michael Lavorgna's picture
Thanks for your kind words especially on the music. This makes me very happy.

This article and the included references were solely concerned with describing how noise can be an issue in digital cables. As you've seen in the attached papers, the quality of the cable matters.

The question of how different cables can sound different is a much more complex discussion for another day. In the mean time, I know of no better way to experience this than to listen.

24bitbob's picture

I'll give you one back (I picked this up from a UK HiFi Publication).
Anouar Brahem, his album Souvenance A Tunisian jazz artist recorded with the Swiss-Italian Orchestra. You can get it on highresaudio or Qobuz in 24/96.
Cables don't matter with music this good. Well worth the few mouse clicks it would take to locate.

Glen65's picture

I work in the Canadian Navy as an electronics tech and we are currently converting/upgrading most of our systems to digital transmission of one sort or another and we are see a huge problem with noise because of all the transmitters, close proximity power generators, and High frequency 400hz lines. not to mention dirty power. its making it very difficult to get clean enough signals that the receiving end can use with out dropped information or so much noise transmitted to the receiving end that it spills over into the monitors of that equipment. We also have issues with IO cards failing prematurely. is a real world problem and I may not go for the big money cables but I can see how a good quality cable will make a difference in transmission of clean signals and to prevent other signals from effecting the source and receiving end. I've seen power supplies put noise on the out put of a source cable and transmit it to the other end on poorly shielded cables for example. there is so much out there to interfere you have no idea how much RF there is out there we use special systems to detect various RF signals and our world is a cloud of RF from so many sources.

monetschemist's picture

Funny to see this today. Earlier this afternoon "it was time" to upgrade my daughter's Dell to Ubuntu 15.04.

A lot of laughter when we were restoring her home directory from the external USB drive and the right speaker was making hooting noises during the restore... I say hooting but it was more like mmm noises from another orifice...

I believe in noise, related to digital stuff coming over the USB. I heard it today, quite loud, in the kitchen.

PeterV's picture

Dear Michael, like many readers I am intrigued by this discussion on RFI noise and jitter influences on audio signals and systems. What I miss though, is a more fundamental view and discussion on the way we are treating our digital information from the beginning. First of all data extraction from a CD starts with a proper reading of the bits inside our CD's. The speed of reading these data and the quality of the optical and servo-drive of the CD player and also the type of CD material itself has influence on the audio quality. Many audiophiles have been colouring the outside of their CD with green or black paint or even cutting of the side ring of the CD with a special device. All to increase the sound quality during the data extraction process from laser to DAC.. I therefore have some concerns what might be happening with our digital data during high-speed data extraction by computer CD-rom platforms... Any CD readout will have a higher or lower Block Error Rate (BLER) or Burst Error (BRST)and any CD drive has its own signature and capability and limitations during this data extraction proces. We cannot afford to solely apply CEC TXLO or 47lab pitracer devices at 1:1 extraction speed for the best optical readout the market has to offer.. This means that many of our .flac,.alac or .wav files are of different quality and they will all contain different readout errors BLER to begin with.. Then the internet started offering us download facilities and we started storing our data on a harddrive of a PC or a NAS. We can download complete albums in various forms, either via direct download from (hopefully..) qualitative sources who gather and supply their digital music files not via optical readout of CD sources, but from the studio master tapes. Many digital data are available as well via torrent facilities and computer technology is able to create a complete album from thousends of truncted internet sources.. . At the end, all these data are stored 'noisefree' in digital form somewhere in our home. So far so discussion whether these data include allready errors due to RFI or jitter or incorrect data extraction. Data on a NAS or USB stick or HD or SSD drive seem to be inert to all forms of errors which potentially have been introduced during the collecting proces...? Or is this wishfull thinking..? Anyway, the last year, many reviews have been written and much debate is going on regarding the last stage of data transport from our PC to DAC via ethernet- or USB-cable or wifi and bluetooth devices. Audiophiles are sensitive to minute changes and improvements of these cables and I have to admit, I am one of them and due to that a bit frustrated as well. I simply do not understand why this is so much audible and why the audio industry is not able to solve this problem fundamentally. What is the largest cause of digital noise (if such a thing exists at all to begin with..) and how can this be prevented or completely neutralised? Why is only this last stage of transport to and conversion in our DAC sensitive to RFI and jitter and not the data which are stored on the internet, NAS, or other sources? Is it so difficult to make a 100% error and noise free transport from storage to DAC..? Why all these debates on computer noise and influences of computer processes during data transmission to DAC as well? Influence of 'competing' clock frequencies between DAC and PC and necessity to develop asynchronous USB connection devices and again conversions etc etc. I do not understand this at all.. Is it impossible to store a complete CD ( or more) on a RAM, isolated from all influences, disturbances, noise etc. and located very close, point-to-point to the DAC conversion chip? Direct i2S-i2S connection, only 1 masterclock which can be operated at various levels, depending on the source either 44.1 or 88.2 or 192 and higher and dsd etc frequencies? Wouldn't this be the most simple and interference free solution for digital storage and playback of music? I suppose this is what happens in a proper music streaming device, but still these devices apply their a buffer chip design which is still sensitive to digitakl noise and are using USB or other cable connections which have influence on the sound quality.. A simple DVD, blue-ray or 4K player, which is processing a much higher amount data, do apply for years a much larger buffering solution. This principle was also applied in some extremely expensive CD players several years ago and was baptised as 'digital lens'.. Audiophiles were impressed.. So how can and should this problem be solved in a streamer-DAC device? I firmly believe that the best cable is no cable at all.. Someone in the audio industry should be able to design an interface between data source and DAC which is 100% robust and insensitive to errors introduced during conversion and transport and re-clocking.. As soon as this is realised, we will start questioning: what is the quality of our data anyway? coming back to the discussion how these have been extracted.. This offers a chance to record companies. Not every .flac or .alac. or .wav file is the same sicne they might contain different read-out errors (or am I wrong with this assumption..?) Maybe I am a bit philosophical, since I have no technical or electronical background at all. I am just waiting for the holy grail of sound :-) So I look forward to your feedback and other readers. Thanks and all the best!

PeterV's picture

Dear Michael,

What is your opinion regarding the distortion influences like RFI and jitter in the digital domain? If these audible influences only occur due to timing errors and other sources, will they be stored as well? If so, then it will be easy to reveal that such distortion will increase by multiple transport, storage, jitter and rfi introduction experiment... If not, then it will reveal that digital (music) data storage is actually a very robust jitter and rfi free technology, agree? And if this is indeed the case, would you also agree that the most robust, jitterless and rfi free transport of data can be achieved by placing the music data physically as close as possible, conversionless point-to-point direct I2s-i2s RAM ( SD card?) to DAC chip applying 1 masterclock will be the most harmonically architecture for digital music playback? What is wrong with this picture, am I overlooking something? Please enlighten me :-)

audioFan23's picture

There is an application note on emoSystems .com which describe audio noise management techniques for streaming audio over ethernet.